The homeobox gene family encodes transcription factors that regulate morphogenesis and cell differentiation during embryogenesis by activating or repressing the expression of target genes (Boersma et al., 1999). In addition, several homeobox genes have recently been shown to be associated with cancers (Raman et al., 2000; Abate-Shen, 2002; Samuel and Naora, 2005; Yoshida et al., 2006). EMX2 is a human homologue of the Drosophila empty spiracles gene (ems), a homeodomain-containing transcription factor with important functions during early development. For example, mice harboring homozygous mutation of EMX2 (EMX2−/−) exhibit small cerebral hemispheres and olfactory bulbs (Dalton et al., 1989). EMX2 affects the proliferation of adult neural stem cells by regulating the frequency of symmetric divisions that generate two stem cells within the adult neural stem cell population, and overexpression of EMX2 decreases the frequency of symmetric divisions (Galli et al., 2002). EMX2 controls mammalian reproduction by adjusting endometrial cell proliferation without affecting differentiation (Taylor and Fei, 2005). Moreover, it has been reported that loss of EMX2 function leads to ectopic WNT1 expression in the developing mammalian telecephalon, resulting in cortical dysplasia (Ligon et al., 2003). The WNT pathway is well known to have important roles in (Klaus and Birchmeier, 2008), including that in lung cancer (Mazieres et al., 2004; You et al., 2004; He et al., 2005; Huang et al., 2008; Akiri et al., 2009). There have been only a limited number of recent studies suggesting possible involvement of EMX2 in human cancer. For example, EMX2 may be anti-proliferative in the endometrium, and its expression is decreased in endometrial tumors (Noonan et al., 2001, 2003). EMX2 also displays methylation but rarely in non-seminomas (Lind et al., 2006). The role of EMX2 in tumorigenesis, however, is still largely unknown. In this study, we seek to investigate the role of EMX2 in human lung cancer.

We first examined the mRNA levels of EMX2 in human lung cancer tissue samples and their matched adjacent normal tissues obtained from 64 patients with lung cancer. Upon comparison, 71.8% (46–64) lung cancer samples analyzed were found to have less EMX2 expression than their matched adjacent normal tissues (Figure 1a), and this downregulation was statistically significant (mean values of EMX2 expression measured by quantitative RT–PCR were 3.78 and 18.01 in cancer tissues and their matched adjacent normal tissues, respectively; P<0.001). Diminished/absent EMX2 expression was consistently associated with hypermethylation of the EMX2 promoter in these cases evaluated by qMSP (Figure 1b showed an example of 10 paired tissue samples). We also analyzed EMX2 expression and the EMX2 promoter methylation status in 12 lung cancer cell lines to verify these results. In all, 10 of the 12 lines examined were found to lack EMX2 expression (Figure 1c). Using qMSP, we found that the EMX2 promoter in the same 10 cells lines was also methylated (Figure 1d). Next, using a de-methylating reagent, 5-aza-2′-deoxycytidine (DAC), we restored EMX2 expression in cell lines with initially silenced EMX2 (Figure 1e). These data indicate that epigenetic modification may be one of the important mechanisms to silence EMX2 gene in lung cancer. Interestingly, methylation silencing was recently reported for several other homeobox gene family members in cancer. For example, HOXA5 was identified as a direct transcription activator of tumor suppressor p53 and HOXA5 silencing by hypermethylation consequently limited p53 expression in breast cancer (Raman et al., 2000). The HOXA9 promoter was found to be frequently methylated in non-seminomatous TGCT (Lind et al., 2006). Our data also revealed that in a few cases, promoter hypermethylation of EMX2 did not correlate with decreased EMX2 expression, suggesting that alternative mechanisms may account for EMX2 downregulation. Indeed, a report showed that EMX2 transcripts were reduced in a subset of endometrial cancers investigated with a 35% incidence of LOH for the 10q25.3–q26.1 region that included the EMX2 gene. Sequencing analysis uncovered multiple EMX2 variants, including somatic mutations and polymorphisms (Noonan et al., 2001).

We next investigated the roles of EMX2 in the growth of lung cancer cells in which EMX2 was methylation silenced. One week after transfection and subsequent G418 selection, we found that EMX2 restoration in H1299 led to significant proliferative suppression (MTS: P<0.001; colony formation assays: P<0.001) without dramatic morphological changes (Figures 2a–c). In contrast, when endogenous EMX2 expression in A427 was silenced by anti-EMX2 shRNAs, cell proliferation was stimulated (MTS: P<0.001; colony formation: P=0.04) with dramatic morphological changes (cells were larger with more branches after EMX2 shRNA treatment) (Figures 2a–c). These results complement and support the data observed in H1299 cells. In addition, by using H1299 stably transfected with EMX2, we examined the chemo-synergistic effect between EMX2 and cisplatin, a widely used chemo drug in the clinic to treat lung cancer (Figure 2d). We observed that the suppressive potential of moderate doses of cisplatin was significantly enhanced by reexpressing EMX2 in H1299 cells (P<0.0001), indicating potential future therapeutic role for EMX2 in combination with current cytotoxic agents in lung cancer.

To further assess the function of EMX2 in lung cancer progression, we performed a cell invasion assay and found that the percent invasion of EMX2-stably transfected H1299 cells was significantly reduced (P<0.05) and that the percent invasion of EMX2 shRNA-stably transfected A427 cells was significantly increased (P<0.05) (Figure 3a). Consistently, using a 3D-spheroid model to mimic an in vivo microenvironment, we observed that EMX2-stably transfected H1299 formed smaller spheroids (P<0.05) and EMX2 shRNA-stably transfected A427 formed larger spheroids (P<0.05) (Figure 3b). More importantly, control (non-EMX2 expressing) H1299 cells formed 3D-spheroids with multiple invasive cellular branch-like structures indicating broken basal membrane. However, EMX2-stably transfected H1299 formed a less invasive rounder phenotype (P<0.001) (Figure 3c). In contrast, A427 cells treated with control shRNA formed more round/less invasive spheroids than those treated with EMX2 shRNA (P<0.001) (Figure 3d). Together, our results suggest that EMX2 may have a role as a novel suppressor of malignant lung cell progression or metastasis. To test this hypothesis and explore mechanisms involved in EMX2 function in lung cancer, we examined the canonical WNT signaling pathway known to be aberrantly activated in lung cancer and important for proliferation, survival and metastasis of lung cancer cells (Mazieres et al., 2004; You et al., 2004; He et al., 2005; Huang et al., 2008; Akiri et al., 2009; Nguyen et al., 2009). EMX2 cDNA and anti-EMX2 shRNA were stably transfected into both H1299 cells lacking EMX2 expression and into EMX2-expressing A427 cells, respectively. After stable transfectants were established, we used real-time PCR to examine EMX2 expression. H1703 cell line expressing endogenous EMX2 at similar levels to that of adult normal lung tissue (Figure 1c) was used as an EMX2 expression level control. We confirmed that EMX2 expression in EMX2-stably transfected H1299 was compatible with the physiological expression in normal lung, and determined that EMX2 levels in shRNA-stably transfected A427 were significantly downregulated (Figure 4a). Interestingly, we observed that EMX2 restoration in H1299 cells led to decreased WNT signaling-dependent transcription activity (Figure 4b). In contrast, shRNA silencing of EMX2 in A427 cells resulted in increased transcription activity of the canonical WNT signaling pathway (Figure 4c). In these stable lines, analysis of the canonical WNT downstream effector cytosolic β-catenin and its target Cyclin D1, corroborated the association between EMX2 and the canonical WNT pathway in lung cancer (Figure 4d). After treatment of H1299 cells with DAC, a de-methylation agent, we found that both cytosolic β-catenin and Cyclin D1 were downregulated, confirming our hypothesis (Figure 4e). After DAC treatment in A427 cells, we also observed some downregulation of these WNT downstream effectors (Figure 4e). Our interpretation is that DAC is not a specific de-methylation agent for a single gene such as EMX2. Thus, expression of other tumor suppressors that are known to be methylation silenced in lung cancer such as WNT pathway antagonists WIF-1 and SFRP (Mazieres et al., 2004; Fukui et al., 2005) should also be re-activated after the DAC treatment. This would contribute to the downregulation of those WNT downstream effectors. Moreover, our microarray analysis revealed that restoration of EMX2 expression in lung cancer cells downregulated metastasis genes such as S100P and S100A4, members of the EF-hand calcium-binding protein family (Figure 4f). These proteins have been reported to induce metastasis in rodent mammary model systems for breast cancer and are apparently associated with poor patient outcome in breast, colon, lung and esophageal carcinomas (Donato, 2001; Heizmann et al., 2002; Zimmer et al., 2003; Marenholz et al., 2004). It is possible that EMX2 regulates transcription of these metastasis genes in lung cancer by directly or indirectly downregulating canonical WNT signaling. This is supported by recent evidence that S100P and S100A4 may be direct downstream targets of canonical WNT signaling (Stein et al., 2006; Ganesan et al., 2008). Taken together, our results suggest that EMX2 may suppress lung cancer, either directly or indirectly, through regulation of the canonical WNT signaling activity. In addition, a preliminary analysis of co-expression of several WNT family genes along with EMX2 in lung cancer cell lines (Supplementary Figure S1) supported a possible association between reduced EMX2 expression and canonical WNT activation in lung cancer. It is possible that EMX2 functions as a transcriptional repressor in a similar manner in lung cancer as it does during embryonic development. However, other co-factors may have important roles in EMX2-related transcription regulation. The identification of direct target(s) of EMX2 and its interacting proteins is needed to further elucidate the function of EMX2 and its relationship with canonical WNT signaling in lung cancer.