H3K9 Histone Methyltransferase G9a Promotes Lung Cancer Invasion and Metastasis by Silencing the Cell Adhesion Molecule Ep-CAM

Aberrant DNA methylation is the primary epigenetic mechanism for regulating gene expression in human cancers (1, 2). Recent advances have shown that these DNA methylation changes are linked with the presence of an aberrant pattern of histone modification (3–7). This histone code (e.g., acetylation, methylation, phosphorylation, ubiquitinylation, and sumoylation), alone or together with DNA methylation, has a pivotal role in organizing nuclear architecture, which, in turn, is involved in regulating transcription. For example, DNA methylation is typically associated with heterochromatin and transcriptionally repressed euchromatic regions (8, 9). Current evidence suggests that H3K9 methylation and silencing of the p16^ink4a tumor suppressor gene can occur before CpG methylation (10). Aberrant silencing of tumor suppressor genes DSC3 and MASPIN in breast epithelial tumor cells has been previously linked to DNA methylation and H3K9 dimethylation of their promoters (11). It has been reported that changes in global levels of individual histone modifications are independently predictive of the clinical outcome of prostate cancer, gastric adenocarcinomas, as well as breast, ovarian, and pancreatic cancers (12–14). These results support the hypothesis that aberrant histone modification patterns are critically involved in the tumorigenic process.

G9a is a recently identified Su(var), Enhancer of Zeste, Trithorax (SET) domain–containing protein with histone lysine methyltransferase activity (15). G9a is a euchromatin-localized histone methyltransferase (HMT) and catalyzes the methylation of histone H3 at lysines 9 and 27 (H3K9 and K27; ref. 16). Targeted deletion of G9a in knockout mice revealed that G9a is predominantly responsible for dimethylation of H3K9 (H3K9me2; ref. 17). G9a plays an important role in the silencing and subsequent de novo DNA methylation of embryonic and germ-line genes during normal development (8) and is necessary for the maintenance of the DNA methylation profile of mammalian cells (9).

The role of HMTs in promoting tumorigenesis and the progression of human cancers has begun to emerge. Highly expressed EZH2 has been observed in metastatic prostate cancer, lymphomas, and aggressive breast cancer (18). SUZ12 is upregulated in some colon, breast, and liver cancers (19). G9a and EZH2 are also upregulated in hepatocellular carcinoma (20). The suppression of either G9a and SUV39H1 reduced cell proliferation and anchorage-independent colony growth while inducing apoptosis in immortalized normal human bronchial epithelial cells (21). Knockdown of G9a and SUV39H1 in PC3 prostate cancer cells inhibited cell growth and led to morphologically senescent cells with telomere abnormalities (22). These studies indicate that G9a seems to be required for the maintenance of the malignant phenotype.

Tissue invasion and metastasis are the major causes of cancer-related death (23). Some studies have found that HMTs specifically affect metastasis. EZH2 is linked to cell proliferation and invasion in prostate cancer and breast cancer (18, 24) and significantly associated with distant metastases in gastric cancer (25). A recent study further established a causal role for EZH2 in driving metastasis in prostate cancer (26). The functional roles of other members of the HMT family such as G9a in cancer remain obscure. Therefore, we investigated whether G9a might also function as a regulator of metastasis. In this report, we explore whether G9a represents a new metastasis promoter within the HMT family. We then define the mechanism by which G9a promotes metastatic lung cancer and show that epigenetic suppression of downstream Ep-CAM is an important mechanism by which G9a triggers metastasis. Furthermore, elevated levels of G9a correlate with poor prognosis and may act as an independent prognostic factor.

The tissues used were from the Cancer Tissue Core of the National Taiwan University Hospital. None of the patients had received preoperative neoadjuvant chemotherapy or radiation therapy. The surgical specimens had been fixed in formalin and embedded in paraffin before they were archived. We used the archived specimens for immunohistochemical staining. The histologic diagnosis of lung adenocarcinoma was made according to the recommendations of WHO. Tumor size, local invasion, lymph node metastasis, and final disease stage were determined as described previously (27). Follow-up of patients was carried out up to 200 months. Patients who died of postoperative complications within 30 days after surgery were excluded from the survival analysis.

A four-point staining intensity scoring system was devised for determining the relative expression of G9a in cancer specimens; the staining intensity score ranged from 0 (no expression) to 3 (maximal expression). The results were classified into two groups according to the intensity and extent of staining: in the low-expression group, either no staining was present (staining intensity score = 0) or positive staining was detected in less than 10% of the cells (staining intensity score = 1), and in the high-expression group, positive immunostaining was present in 10% to 30% (staining intensity score = 2) or more than 30% of the cells (staining intensity score = 3). All of the immunohistochemical staining results were reviewed and scored independently by two pathologists.

The antibodies included anti-human Ep-CAM (Cell Signaling Technology) and anti-G9a (R&D Systems, Perseus Proteomics). Immunodetection was performed with an EnVision dual link system-HRP detection kit (DAKO Corporation).

Lung cancer cell lines were grown in RPMI 1640 plus 10% fetal bovine serum (Invitrogen/Gibco) in a humidified atmosphere containing 5% CO₂ at 37°C. Lung adenocarcinoma cell lines (CL1-0 and CL1-5) were established in the National Health Research Institutes laboratory and displayed progressively increasing invasiveness (28). Other lung cancer cell lines (PC14, H441, and H1299) were obtained from the American Type Culture Collection.

The lentiviral G9a shRNA constructs were purchased from the National RNAi Core Facility in Academic Sinica, Taipei, Taiwan. The Ep-CAM shRNA constructs were obtained from Open Biosystems. The target sequences of these shRNA are described in Supplementary Table S1. Lentiviruses were produced by cotransfecting shRNA-expressing vector and pMD2.G and psPAX2 constructs into 293T cells by using calcium phosphate. Viral supernatants were harvested, titered, and used to infect CL1-5 or H1299 cells with 8 μg/mL polybrene. Cells were selected using 2 μg/mL puromycin. Luciferase-expressing cells (CL1-0/Luc or CL1-5/Luc) were established by infecting with lentivirus-expressing pWPI-Luc-ires-GFP vector. CL1-0/HA-G9a cells were established by infecting with lentivirus-expressing pWPXL-HA-G9a vector.

Invasion and migration assays were performed as published (ref. 30; see Supplementary Data).

Western blot analysis was performed with the primary antibodies anti-G9a, anti-H3K9me2 (Upstate), anti–Ep-CAM (Cell Signaling Technology), and anti–α-tubulin (Sigma-Aldrich).

All animal works were done in accordance with a protocol approved by the National Taiwan University College of Medicine and National Taiwan University College of Public Health institutional animal care and use committees. Age-matched nonobese diabetic severe combined immunodeficient (SCID) female mice (6–8 weeks old) were used. For experimental metastasis assays, 1 × 10⁶ cells were resuspended in 0.1 mL of PBS and injected into the lateral tail vein. Lung metastatic progression was monitored and quantified using the noninvasive bioluminescence system (IVIS-Spectrum). For orthotopic metastasis assays, cells (1 × 10⁶ CL1-0 cells, 5 × 10⁵ CL1-5 cells) were resuspended in a 1:1 mixture of PBS and GFR-Matrigel (BD Labware). This mixture was then injected into the left lateral thorax of each mouse as previously described (29). Metastatic nodules in the right lung were quantified using a dissecting microscope at each end point.

The Ep-CAM promoter (−250 to +90) was used as described (30). The Dual Luciferase Reporter assay system (Promega) was used with TK-Renilla luciferase plasmid as a transfection efficiency normalization control.

Chromatin immunoprecipitation (ChIP) assays were performed according to the manufacturer's protocol (Upstate). The chromatins were incubated with 4 μg of anti-K9 dimethylated histone H3, anti–acetylated histone H3, anti-G9a (Upstate), anti-DNMT1, anti-Sp1, anti-HDAC1, and anti-P300 antibody (Santa Cruz Biotechnology) at 4°C overnight. Immunoprecipitated DNA was analyzed by quantitative PCR by using specific primers as described in Supplementary Table S2.

DNA was treated with bisulfite and purified for PCR as described previously (31). Primer sequences were as described in Supplementary Table S2. The sequences of the Ep-CAM promoter primers were 5-AAGGAAGTTTTAGTATAGAATTTTTAAATT-3 (F) and 5-AAAAAAATAAATAAACTCCCCTCC-3 (R). The PCR products were ligated into pGEM-T vector (Promega) and transformed into DH5α. Plasmid DNA was isolated and then subjected to sequence analysis.

Tissues were processed by fixing in 4% buffered formalin and then embedding in paraffin wax. Sections (3 μm) were stained with H&E for histopathologic analysis.

All observations were confirmed by at least three independent experiments. The data were presented as mean ± SD. ANOVA was used to evaluate the statistical significance of the mean values. Cox proportional hazards regression was used to test the prognostic significance of factors in univariate and multivariate models. Spearman's rank correlations were determined for comparison of G9a and Ep-CAM immunostaining. All statistical tests were two-sided, and P < 0.05 was considered significant.

Comparisons of the expression levels of G9a in tumor tissues and control normal tissues were made. Immunohistochemical examination of 32 paired lung adenocarcinoma specimens revealed a significantly higher expression of G9a in tumor tissues (P < 0.0001; Supplementary Fig. S1). Similar results were also observed from the analysis of 22 paired lung squamous cell carcinoma specimens (P < 0.0001; Supplementary Fig. S1). Collectively, G9a was expressed at a lower level in normal lung tissues and preferentially expressed in lung tumor tissues.

The prognostic significance of G9a expression was determined by assessing its nuclear staining using 160 human lung cancer specimens with known clinical follow-up records. Figure 1A shows representative examples with different G9a scores. The relationships between the levels of G9a expression and the clinicopathologic characteristics of lung cancer are summarized in Supplementary Table S3. Among these specimens, we found that high G9a expression level (scores of 2 and 3) correlated strongly with reduced overall survival relative to tumors with low G9a expression level (scores of 0 and 1) as shown in Fig. 1B. Similar results were obtained for disease-free interval (Fig. 1C). The prognostic significance of G9a was also performed by using tissue microarray (TMA) containing 119 cases from an independent lung cancer patient cohort. Similar results were observed (Supplementary Fig. S2A; Supplementary Table S4). In a multivariate Cox model including G9a score, histology type, tumor stage, tumor status, lymph node involvement, and metastasis, G9a levels and metastasis were found to be significant predictors of outcome (Table 1; Supplementary Fig. S2B). Taken together, our data indicate that higher levels of G9a predict poor prognosis in lung cancer.

To elucidate a link between G9a expression and tumor cell invasiveness, we used a set of lung adenocarcinoma cell lines (CL1-0 and CL1-5) that were designed to exhibit progressive invasiveness abilities as previously described (28). Western blot analysis showed that G9a protein levels were significantly elevated in the highly invasive CL1-5 lung cancer cells as compared with the poorly invasive parental CL1-0 cells (Fig. 2A, left). Next, we asked whether a correlation between G9a expression and tumor cell invasiveness occurred in other cell lines. We observed abundant G9a expression in the highly invasive H1299 lung cancer cell line, but low G9a expression in poorly invasive lung cancer cell lines (PC14 and H441; Fig. 2A, right). G9a and GLP have been described to form a homomeric or heteromeric complex through their SET domain interaction, both of which are crucial for H3K9 methylation of euchromatin (32). We also examined GLP expression and global H3K9me2 in these lung cancer cell lines. Our results showed similar levels of GLP expression in all cell lines tested but different levels of G9a expression (Fig. 2A). Global H3K9me2 status was not completely correlated with G9a expression in lung cancer cell lines (Fig. 2A).

To determine whether G9a modulates tumor cell migration and invasion, we knocked down G9a in CL1-5 and H1299 cells using G9a-specific shRNAs. We initially determined the influence of G9a knockdown on cell proliferation, apoptosis, and senescence as previously described (21, 22). Our results showed that G9a-knockdown cells exhibited no significant differences in proliferation, cell cycle, and senescence profiles as compared with controls (Supplementary Fig. S3A–D).

Interestingly, expression of two G9a-specific shRNAs significantly reduced the mRNA and protein levels of G9a with concomitant inhibition of the migration and invasion potentials of CL1-5 and H1299 cells (Fig. 2B; Supplementary Fig. S4). Overexpression of G9a in poorly invasive CL1-0 cells was also found to enhance the migratory and invasive abilities of CL1-0 cells (Fig. 2B, right). G9a knockdown in A549 cells with lower endogenous G9a expression was also found to significantly inhibit their migration abilities (Supplementary Fig. S5). Taken together, these results indicate that G9a regulates the migration and invasion of lung cancer cells.

To determine whether the enzymatic activity of G9a was required for its effect in promoting cancer cell motility and invasion, CL1-5 and H1299 cells were transfected with a dominant-negative mutant of G9a (DN-G9a) containing two amino acid substitutions within the catalytic domain (N903H and L904E), which abolishes methyltransferase activity (33, 34). Figure 2C shows that DN-G9a transfection significantly reduced H3K9me2 level in CL1-5 and H1299 cells and inhibited their migratory and invasive abilities. On the basis of these results, we suggest that the HMT activity of G9a is required for the migratory and invasive phenotype of lung cancer cells.

To evaluate the role of G9a during metastasis, we used an experimental metastasis model in which cancer cells were i.v. injected into the lateral tail veins of mice. As shown in Fig. 3A, we stably knocked down G9a in CL1-5/Luc cells, which stably expressed luciferase, and implanted these cells i.v. into SCID mice (Fig. 3A). Four weeks after injection, the lung was removed and metastasis was monitored by bioluminescence imaging. G9a knockdown resulted in less detectable lung metastases compared with controls (Fig. 3A, right). Quantification of the metastatic nodules present using dissecting microscopy and histologic analyses of the lung dissected from each mouse confirmed that the number of lung metastases was drastically reduced in mice carrying G9a-knockdown tumors (Fig. 3B).

We next investigated whether ectopic expression of G9a is sufficient to induce invasive activity in low-metastatic cancer cells using an orthotopic lung cancer model as described (29). Low-metastatic CL1-0 cells stably expressing G9a were orthotopically injected into the upper lobe of left lung; tumor growth and metastasis were monitored and quantified by bioluminescence imaging. Tumor metastasis into the right lung was increased in the G9a-overexpressing group (Supplementary Fig. S6A). Quantification of metastatic nodules confirmed that right lung metastasis was significantly increased in mice carrying CL1-0/HA-G9a tumors (Fig. 3C). Overexpression of G9a did not alter in vivo tumor growth rate (Supplementary Fig. S6B). Furthermore, G9a-overexpressing CL1-0 cells were found to invade into mediastinal lymph nodes (Fig. 3D). Hence, G9a function is essential and sufficient to promote lung cancer metastasis.

Microarray RNA expression profiles were compared between CL1-5 cells with control shRNA and G9a shRNA2 to identify invasion/migration-related genes directly regulated by G9a. Reverse transcription-PCR analysis was used to further validate the expression levels of these genes (Supplementary Fig. S7A). Our results showed that the mRNA levels of E-cadherin and Ep-CAM were significantly upregulated in G9a-knockdown cells compared with control (Supplementary Fig. S7A). E-cadherin and Ep-CAM protein levels were then determined in G9a-knockdown CL1-5 and H1299 cells. G9a knockdown significantly increased Ep-CAM protein levels in CL1-5 (7.5- and 16-fold) and H1299 cells (2.4- and 3.6-fold; Fig. 4A), whereas endogenous E-cadherin protein levels were extremely low in both cell lines. G9a knockdown induced low level of E-cadherin expression in CL1-5 but not in H1299 cells (Fig. 4A). Microarray RNA expression profiles were also determined in G9a-overexpressing CL1-0 cells and vector control. The results showed inverse correlation of RNA expression with G9a knockdown (Supplementary Fig. S7A). Western blot analysis further confirmed that ectopic overexpression of G9a induced a more than 2-fold reduction of 3Ep-CAM protein level in CL1-0 cells (Supplementary Fig. S7B).

We next investigated whether G9a expression inversely correlated with Ep-CAM levels in human lung cancer patients. Immunohistochemistry analysis of lung cancer specimens revealed an inverse correlation between G9a and Ep-CAM expression (tested by Spearman's nonparametric correlation test, correlation coefficient = −0.4, P < 0.05; Supplementary Table S5). The representative immunohistochemical staining for Ep-CAM and G9a on serial sections revealed inverse staining patterns in lung adenocarcinoma tissues (Fig. 4B). To extend the inverse correlation between G9a and Ep-CAM in human lung cancer, we further determined Ep-CAM levels in serial lung cancer TMA set in which G9a was examined. The results also revealed an inverse correlation between G9a and Ep-CAM expression (tested by Spearman's nonparametric correlation test, correlation coefficient = −0.189, P < 0.05; Supplementary Fig. S8A). The correlation of high G9a expression with low Ep-CAM expression in human lung cancer patients is consistent with our finding that knockdown of G9a can upregulate Ep-CAM in lung cancer cells.

These observations led us to examine whether Ep-CAM could reverse G9a-mediated phenotypes. As shown in Fig. 4C and Supplementary Fig. S9, infection and expression of two shRNAs significantly reduced the levels of Ep-CAM mRNA and protein with concomitant increase in the invasive ability in vitro of CL1-5 cells that were prior infected with G9a shRNA compared with control (Fig. 4C). To investigate whether Ep-CAM knockdown could reverse the in vivo metastatic phenotypes attributed to G9a functions, we orthotopically injected mice with CL1-5 cells expressing combinations of G9a shRNA, Ep-CAM shRNA, and control vector. G9a knockdown showed more than 70% reductions in total luciferase counts (P < 0.05) in orthotopic tumors (Supplementary Fig. S10). Ep-CAM shRNA, however, could abate primary tumor growth reduction in the presence of G9a shRNA (Supplementary Fig. S10). Ep-CAM knockdown was found to restore lung metastasis in G9a shRNA2–expressing CL1-5 cells to 77% of control levels in orthotopic left to right lung metastasis assay (Figs. 3C and 4D). Taken together, these data indicate that the ability of G9a shRNA to inhibit metastasis is attributable, in significant part, to its capacity to upregulate Ep-CAM.

To further characterize the mechanism of G9a-mediated downregulation of Ep-CAM expression, we next examined G9a binding and H3K9 dimethylation at different regions of the human Ep-CAM gene by using ChIP analysis with antibodies against G9a and dimethyl-H3K9 (H3K9me2). Four representative regions spanning ∼2.5 kb upstream of the transcription initiation site of the Ep-CAM gene were investigated (Fig. 5A, top). The results showed that G9a and H3K9me2 were found in region P3 only in CL1-5 cells (Fig. 5A). This region also contains a consensus binding site for Sp1, a transcription factor that has been reported to regulate Ep-CAM transcription activity (35). To delineate the role of Sp1 in G9a knockdown–mediated Ep-CAM transactivation, we transfected CL1-5 cells with shRNAs against Sp1 that have been previously infected with either Luc or G9a shRNA. Figure 5B shows that G9a knockdown resulted in a significant induction of the Ep-CAM promoter (−250/+90) luciferase activity (lane 2 versus lane 4), which was diminished in the presence of mithramycin A (a steric inhibitor that prevents Sp1 binding to DNA) or Sp1 shRNA, but not in the presence of P50 shRNA (lanes 4–9; Fig. 5B). These results suggest that G9a-regulated Ep-CAM gene transactivation is dependent on an endogenous Sp1 transcription factor.

To elucidate the assembly of the repressor protein complex that binds to the Ep-CAM promoter, we performed ChIP assay of the Ep-CAM promoter in G9a-knockdown CL1-5 cells and control. As expected, loss of G9a resulted in loss of H3K9me2, HP1, DNMT1, and HDAC1 binding at region P3 compared with SP1 (Fig. 5C). Methylation-specific PCR (MS-PCR) and bisulfite sequencing were further performed to evaluate whether G9a-knockdown is responsible for CpG demethylation of the Ep-CAM promoter. As shown in Fig. 5D, a clear unmethylated band of the Ep-CAM promoter was observed in G9a shRNA2–expressing CL1-5 cells by MS-PCR analysis (Fig. 5D, top). Bisulfite sequencing results confirmed that G9a knockdown significantly reduced Ep-CAM promoter methylation (Fig. 5D, bottom row). These results indicate that both DNA and histone methylation, along with repressive complexes, mediate Ep-CAM gene repression.

HMT can act as a driver of metastasis (26). Here, we propose for the first time that G9a acts as a promoter of metastasis in the context of lung cancer. G9a is a major HMT that maintains global H3K9me2. Several studies have shown that lower global levels of H3K9me2 predict poor prognosis in prostate and kidney cancers (36, 37). Therefore, we analyzed the correlation between G9a and H3K9me2 level in human lung cancer TMA. We did not observe any significant correlation of G9a expression and global H3K9me2 level (Supplementary Fig. S12A). Global levels of H3K9me2 also did not predict prognosis in lung cancer (Supplementary Fig. S12B). We also found that global H3K9me2 level was not correlated with G9a expression in lung cancer cell lines (Fig. 2A). Global H3K9me2 is dynamic, and it changes as a result of the effects G9a, GLP, and H3K9 demethylases, such as JMJD2A (38), JMJD2C (39), and LSD1 (40), but the complex interplay of these enzymes is poorly understood. Based on our studies, G9a level is a more rigorous marker for cell invasion and prognosis factor in lung cancer. This may be due to the G9a suppression of transcription by independently inducing both H3K9 and DNA methylation (41).

It is reported that G9a and GLP form a stoichiometric heteromeric complex in vivo and function cooperatively rather than redundantly to mediate H3K9 dimethylation at euchromatin (32, 42). Their interaction also stabilizes G9a protein from degradation. We tested the possibility that downregulation of GLP results in a similar phenotype as that observed with downregulation of G9a. We found that knockdown of GLP resulted in migration defect, as in the case of knockdown of G9a (Supplementary Fig. S11). Similar to the result of a previous report (17), GLP depletion also caused G9a protein depletion (Supplementary Fig. S13). Depletion of each protein will affect the function of the G9a-GLP complex. Collectively, our results indicate that downregulation of GLP results in a similar phenotype as that observed with downregulation of G9a. The results also suggest that GLP is essential in the promotion of cell migration by G9a.

In this study, we have provided substantial evidence linking G9a expression with lung cancer progression; however, the possible mechanisms of G9a overexpression associated with lung cancer metastasis remain elusive. Our preliminary data suggest that posttranslational regulation mechanisms may be involved. We have found that G9a expression is higher in lung tumor tissues compared with matched normal tissues (Supplementary Fig. S1). Local invasion is one of the early events for tumor metastasis. We have provided evidence to show that expression of G9a increases cell invasion (Fig. 2B). Taken together, increased expression of G9a may occur as an early event in metastasis processes. Other possibilities may also exist and cannot be ruled out at this moment. Recently, cancer stem cell is believed to act like a seed to metastasize at distant organ on entry into circulation. Indeed, we also found that G9a expression was higher in colon cancer stem cells (sorting against multiple colon stem cell markers such as CD133 and CD44), although this was not observed in our lung cancer cell models. Therefore, it is possible that high G9a expressing cells may contain more cancer stem cell potentials with higher capability for metastasis. Currently, the association between G9a expression and cancer stem cells is under investigation in our laboratory.

Our results showed that knockdown of Ep-CAM partially reversed metastasis suppression due to G9a knockdown in vivo. Furthermore, we showed that patients with high expression of G9a and with concomitantly low Ep-CAM levels had significantly shorter survival time in lung cancer TMA (Supplementary Fig. S8B). These findings are quite interesting and warrant further investigation because Ep-CAM is only one member of a large cohort of metastasis-relevant genes that may be repressed by G9a. Ep-CAM is a 40-kDa epithelial transmembrane glycoprotein that is abundantly present in most epithelial tissues and functions as a homophilic Ca²⁺-independent cell-cell adhesion molecule (43). Loss of membranous Ep-CAM is associated with nuclear β-catenin localization and contributes to reduced cell-cell adhesions, increased migratory potential, and tumor budding (44). Nuclear translocation of β-catenin may cause activation of genes that are regulated by β-catenin. These studies may partially explain why Ep-CAM induces the reversal of G9a-mediated phenotypes. This supports the hypothesis that G9a may act via the pleiotropic regulation of multiple effectors.

In conclusions, we show that G9a is endowed with methyltransferase activity to concomitantly repress the downstream effector Ep-CAM, thereby promoting the invasion step of the invasion-metastasis cascade. Moreover, G9a levels correlate with reduced overall survival and disease-free interval, potentially representing an independent prognostic factor.

No potential conflicts of interest were disclosed.

We thank Dr. Marianne Rots (University of Groningen) for the P39E plasmid and Dr. Kenneth L. Wright (University of South Florida) for DN-G9a and HA-G9a plasmids.

Grant Support: National Science Council, Taiwan (NSC97-2323-B-002-001) and Department of Health, Taiwan (DOH99-TD-C-111-004).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.