Connexins (Cxs), the gap junction proteins, have been found to be downregulated in many types of cancers including gliomas. By restoring gap junctional communication in cancer cell models, the neoplastic phenotype can be reversed, suggesting Cxs are tumor suppressors. Pannexin2 (Panx2) is a member of the novel gap junction protein family, Panxs, and it has been proposed as a brain-specific protein. Recently, gene array analysis showed an overall reduction of Panx2 in gliomas, and a direct correlation was observed between Panx2 expression and post-diagnosis survival in patients. In this study, we explored the potential inverse correlation between Panx2 and glioma oncogenicity. A decrease or absence of Panx2 expression in a panel of human glioma cell lines was found, whereas an appreciable amount of Panx2 was detected in both human brain and astrocytes. Stable Panx2 expression revealed a flattened morphology and increased cell–cell contacts in rat C6 glioma cells similar to Panx1. However, in contrast to Panx1 and Panx3, Panx2 was predominately detected in the cytoplasm in vesicle-like patterns but not at the plasma membrane. Coexpression of Panx2 and Panx1 did not show colocalization of both Panxs. Strikingly, restoration of Panx2 expression significantly reduced in vitro oncogenicity parameters, including monolayer saturation density and anchorage-independent growth, as well as in vivo tumor growth. This study suggests a role of aberrant Panx2 expression during gliomagenesis, and that Panx2 independently functions as a negative growth regulator without Panx1.
Gap junctions are membrane channels that connect the cytosols of adjacent cells, allowing the direct passage of small molecules <1200 Da (Loewenstein, 1987). A lack of cell–cell communication has been shown in neoplasms including gliomas (Mesnil, 2002; Naus, 2002), and restoration of gap junctional intercellular communication in cancer cell models has been found to reverse the transformed phenotypes and to reduce proliferation (Naus et al., 2005). Later studies have also shown a tumor-suppressive role of connexins (Cxs), the canonical gap junction proteins, through protein–protein interactions, further supporting its function as a tumor suppressor (Trosko and Ruch, 2002; Naus et al., 2005).
Pannexins (Panxs) are a novel family of gap junction proteins recently identified as the mammalian homologs of the invertebrate gap junction proteins, innexins. To date, three Panx members have been identified in vertebrates: Panx1, Panx2 and Panx3. A functional role of Panx1 has been shown in intracellular Ca2+ release, propagation of Ca2+ waves in blood vessel endothelial cells, inflammatory response, ischemic cell death and taste cell signaling, advocating the importance of Panx1 under both normal and pathological conditions (Locovei et al., 2006; Pelegrin and Surprenant, 2006, 2007; Thompson et al., 2006; Huang et al., 2007; Romanov et al., 2007). In addition, we showed a reversion of transformed phenotypes of C6 gliomas when Panx1 was overexpressed (Lai et al., 2007), suggesting a tumor-suppressive role of Panxs in neoplasms. However, Panx2 function remains largely undiscovered, and its role in neoplasm has not been examined.
Panx2 transcripts, unlike Panx1, are particularly abundant in the brain in rodents and presumably brain specific in humans (Bruzzone et al., 2003; Baranova et al., 2004; Vogt et al., 2005; Ray et al., 2006). In parallel to the loss of Panx2 transcript in C6 gliomas when compared with its normal counterpart, rat primary astrocytes (Lai et al., 2007), high-throughput cDNA microarray analysis of human brain tumor samples has shown an overall reduction of Panx2 gene expression in gliomas (Litvin et al., 2006). A direct correlation was also found between Panx2 upregulation and post-diagnosis survival in patients with glial tumors using the brain cancer gene expression database REMBRANDT (Repository of Molecular Brain Neoplasia Data, http://rembrandt.nci.nih.gov/rembrandt) (Litvin et al., 2006). Moreover, human Panx2 is located within chromosomal region 22q13.3, where deletions are often found in human astrocytomas and ependymomas (Ino et al., 1999; Oskam et al., 2000; Hu et al., 2004; Ray et al., 2006). Thus, Panx2 expression may be commonly altered during glioma development. Altogether, the findings suggest a role of aberrant Panx2 in gliomagenesis and oncogenicity.
Results and discussion
Reduced Panx2 expression in human glioma cells
To investigate whether aberrant Panx2 expression occurs in gliomas, a panel of human glioma cell lines was examined using western blot analysis and showed a reduction or absence of Panx2 protein when compared with human astrocytes (Figure 1a). The observation is in agreement with the result from gene array analysis of human brain tumor samples in which gliomas have an overall decrease of Panx2 expression (Litvin et al., 2006). A notable level of Panx2 expression was detected in the human brain, advocating a role of Panx2 in the brain (Vogt et al., 2005; Ray et al., 2006; Zappala et al., 2007). An appreciable amount of Panx2 protein was also found in human astrocytes. This is in contrast to the observation made in rats in which Panx2 was only detected in the neurons and reactive astrocytes, but not in normal astrocytes (Vogt et al., 2005; Zappala et al., 2007). The disparity could be a result of species differences and warrants further investigation to determine the type(s) of astrocytes expressing Panx2 in humans. Overall, this finding suggests an inverse correlation between Panx2 expression and glioma oncogenicity.
Panx2 expression alters cellular morphology
Conversely, we investigated whether restoring Panx2 expression could suppress glioma growth. Using a well-established glioma cell line, rat C6 cells, we investigated the effect of Panx2 expression on the transformed phenotype. C6 cells are considered to be of astrocytic origin and poorly coupled by gap junctions because of their limited Cx43 expression (Naus et al., 1991; Barth 1998). In addition, a recent study showed most of C6 cells as cancer stem cells with clonogenic, self-renewal and tumorigenic capacities (Zheng et al., 2007). Furthermore, our recent report determined C6 gliomas to be devoid of Panx transcripts (Lai et al., 2007), which together makes it an ideal model system to study the effect of Panx2 expression on glioma oncogenicity. Several clones with varying expression levels were successfully generated for Panx2-HA and Panx2-EGFP stable transfectants to study the potential correlation between Panx2 and oncogenicity (Figure 1b and c). Clones for the control transfectants, EGFP, were previously created at the same time with tagged Panx2 transfectants (Lai et al., 2007). Endoplasmic reticulum stress is an artifact commonly observed in the overexpression system, perturbing cell growth and survival (Lee, 2005). To address this possibility, stable control- and Panx transfectants were examined for glucose-regulated protein (GRP78/BiP), an endoplasmic reticulum stress marker, and showed no evident increase in expression (Lee, 2005) (Supplementary Figure 1a). On stable transfection, a noteworthy flattened morphology was observed in Panx2-expressing cells (Figure 1d). Interestingly, the flattened morphology was moderate when compared with that of Panx1-expressing cells (Figure 1d). Moreover, Panx2-EGFP signal was observed in the cytoplasm, which is distinctly different from predominately plasma membrane-located Panx1 (Lai et al., 2007).
Panx2 is not associated with the plasma membrane
To further examine the intracellular location of Panx2, a crude membrane fractionation assay identified Panx2-HA and Panx2-EGFP only in the membrane-enriched fraction, suggesting that Panx2 is in membrane-bound vesicles and/or organelles (Figure 1e). We then stained the cell membranes with wheat germ agglutinin, and detected Panx2-EGFP predominately around the perinuclear region in a vesicle-like pattern, but not at the plasma membrane nor at areas of cell–cell contact (Figure 2a). Using specific markers, Panx2 showed a prominent signal overlap with the endoplasmic reticulum but only limited with the Golgi apparatus (Supplementary Figure 1b). Importantly, Panx2 did not appear to aggregate in either organelles, indicating that its trafficking was not encumbered as a result of overexpression; previous studies have shown an aberration in Cx trafficking because of its overexpression (Skerrett et al., 2004; Das et al., 2005; Thomas et al., 2005). One possible explanation for Panx2 distinctive intracellular localization is that it was actively trafficked to the plasma membrane but promptly internalized for degradation, possibly through the endolysosomal pathway previously shown for Cx43 (Laing et al., 1997; Qin et al., 2003; Berthoud et al., 2004). However, Panx2 did not appear colocalized to early endosomes and lysosomes, showing that Panx2 was not being actively degraded in the cytoplasm (Supplementary Figure 1c). It was also unlikely that Panx2 was being actively degraded by the proteasomal degradation pathway, as it was not readily detected in the cytosol-enriched fraction (Figure 1e). Several recent studies have reported Cx40 and Cx43 located in the mitochondria of cardiomyocytes, resembling Panx2 subcellular pattern (Li et al., 2002; Boengler et al., 2005; Schulz et al., 2007). However, co-labeling of mitochondria and Panx2 showed no apparent colocalization (Supplementary Figure 2). Importantly, recent studies have shown a similar expression pattern of endogenous Panx2 in neuronal cells, reactive astrocytes after hypoxia, as well as in spiral ganglion neurons of the cochlea, where Panx2 was detected in the cytoplasm and especially prominent at the perinuclear region (Zappala et al., 2007; Wang et al., 2008). Together, the current finding indicates Panx2 is primarily located within the cytosol and not at the plasma membrane in C6 cells.
Panx2 does not colocalize with Panx1 at the plasma membrane
The subcellular localization pattern of Panx2 is consistent with the finding that Panx2 alone was not able to form either hemichannels or intercellular channels in the Xenopus oocyte system on the cell surface (Bruzzone et al., 2003). Thus, we speculated that Panx2 can be recruited to the plasma membrane by Panx1 as previously suggested to elicit its function (Bruzzone et al., 2003, 2005). Yet, when the two Panxs (tagged or untagged) were coexpressed, while Panx1 was found at the plasma membrane as previously reported (Lai et al., 2007), Panx2 localization remained unchanged (Figure 2b). The difference in findings suggests that Panx1/2 channel formation is not present in C6 gliomas. Collectively, our findings suggest that Panx2 exists in an individual entity, likely in the form of membrane-bound vesicles.
Panx2 reduces oncogenicity of C6 gliomas
Next, we set to determine the effect of restoring Panx2 expression on the oncogenicity parameters in C6 gliomas. Using a growth curve assay, a significant retardation in proliferative rate was discovered in the stable Panx2 transfectants (Figure 3a). As no significant increase in apoptotic nuclei was observed in cells seeded on coverslips in parallel to the growth curve assay (data not shown), the growth reduction by Panx2 is proposed to occur through suppression of the cell cycle, but not cell death. In addition, Panx2 elicited a consistent reduction in saturation density (Figure 3a), which coincides with the flattened morphology of Panx2 transfectants; fewer cells of increased size are required to grow to confluence. A similar growth suppression with human glioma cells was also observed (Supplementary Figure 3). As Panx2 was only found within the cells but not at the plasma membrane like Panx1, and given that further examination also showed that Panx2 does not affect expression of Panx1, Cx43, as well as gap junctional intercellular communication (Supplementary Figure 4), Panx2 may induce its phenotype through a pathway distinct from Panx1. The notion is supported by the transwell assay in which Panx2, in contrast to Panx1, showed no clear effect on cell motility (Figure 3b). Anchorage-independent growth is another hallmark of cancer and can be assessed by the formation of colonies in soft agar. It is noted that both Panx2-HA and Panx2-EGFP transfectants drastically suppressed colony formation efficiency and anchorage-independent growth (Figure 4a and b). This suggests that Panx2 functions in a channel-independent approach. The proposition is strengthened by the soft agar assay as no cell–cell communication was present in the initial phase of the experiment when single cells were seeded in the soft agar. To further examine whether Panx2 elicits tumor-suppressive effects in vivo, tumorigenicity assay was conducted and both Panx2-expressing cells exhibited a strikingly reduced average tumor size when compared with the control. Findings reported here suggest Panx2 expression restores anchorage dependence of C6 gliomas on tethering to the extracellular matrix (ECM). Whether this dependence is through an alteration of the expression of the cell surface receptors, integrins, glioma-derived ECM (that is, vitronectin, tenascin-C), and/or the constitutive activity of survival pathways (that is, PI3K/Akt, MEK/ERK, nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB)), remains to be elucidated (D'Abaco and Kaye, 2007; Chiarugi and Giannoni, 2008).
In summary, the unique subcellular location of Panx2 suggests its functions through pathway(s) independent from gap junction channels. As Panx2 has a 358 amino acid C-terminal tail, which is more than twice the size of that of Cx43 (151 amino acid), we envision that the long Panx2 C-terminal tail may be advantageous in providing potential sites for intracellular protein–protein interaction to mediate Panx2 functions. This study provides the first report on the tumor-suppressive effects of Panx2 and implies Panx2 also functions in glioma stem cells, as a recent study showed that C6 cell line was mainly composed of cancer stem cells with perpetual self-renewal ability (Zheng et al., 2007). In conclusion, Panx2 may serve as a prospective therapeutic target against glial tumors and warrant further investigation.
Conflict of Interest
The authors declare no conflict of interest.
The research was supported by the Canadian Institutes of Health Research. The authors thank Dr WC Sin for the critical review of the paper, Dr L Matsuuchi for the insightful data discussions, as well as Lynne Bechberger for proofreading the paper. CCN is a recipient of a Canada Research Chair.
About this article
Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)
Cell and Tissue Research (2015)