Exogenous overexpression of the metastasis suppressor gene Nm23-H1 reduces the metastatic potential of multiple types of cancer cells and suppresses in vitro tumor cell motility and invasion. Mutational analysis of Nm23-H1 revealed that substitution mutants P96S and S120G did not inhibit motility and invasion. To elucidate the molecular mechanism of Nm23-H1 motility suppression, expression microarray analysis of an MDA-MB-435 cancer cell line overexpressing wild-type Nm23-H1 was done and cross-compared with expression profiles from lines expressing the P96S and S120G mutants. Nine genes, MET, PTN, SMO, FZD1, L1CAM, MMP2, NETO2, CTGF, and EDG2, were down-regulated by wild-type but not by mutant Nm23-H1 expression. Reduced expression of these genes coincident with elevated Nm23-H1 expression was observed in human breast tumor cohorts, a panel of breast carcinoma cell lines, and hepatocellular carcinomas from control versus Nm23-M1 knockout mice. The functional significance of the down-regulated genes was assessed by transfection and in vitro motility assays. Only EDG2 overexpression significantly restored motility to Nm23-H1–suppressed cancer cells, enhancing motility by 60-fold in these cells. In addition, silencing EDG2 expression with small interfering RNA reduced the motile phenotype of metastatic breast cancer cells. These data suggest that Nm23-H1 suppresses metastasis, at least in part, through down-regulation of EDG2 expression. [Cancer Res 2007;67(15):7238–46]
- Suppressors of metastasis
- Cell motility and migration
- Breast cancer
Nm23-H1 was discovered by its reduced expression in highly metastatic murine melanoma cell lines compared with congenic, nonmetastatic lines ( 1). Exogenous expression of Nm23-H1 has been shown to reduce metastasis in 11 independent models, which include melanoma and carcinomas of the breast, ovary, oral cavity, and colon ( 2– 12). According to its definition as a metastasis suppressor gene, Nm23-H1 expression reduced the incidence of metastases without affecting primary tumor size in these models. The in vivo effect of Nm23 on the metastatic process was recently bolstered by the characterization of mice lacking the Nm23-M1 gene, the murine homologue of Nm23-H1 ( 13). The incidence of lung metastases was significantly higher for the Nm23-M1 knockout mice compared with Nm23-M1+/+ mice, when the mice were induced to develop hepatocellular carcinoma (HCC; ref. 14).
Based on its activity in in vitro assays of aspects of metastasis, Nm23-H1 presumably thwarts the spread of cancer at multiple steps along the metastatic cascade. Overexpression of Nm23-H1 nearly abolishes cell motility of MDA-MB-435 cancer cells to multiple chemoattractants including serum, insulin-like growth factor (IGF), platelet-derived growth factor (PDGF), autotaxin, and lysophosphatidic acid (LPA) and reduces their anchorage-independent growth and invasion ( 3, 15, 16). In addition, Nm23-H1 overexpression in MDA-MB-435 cells stimulated aspects of breast differentiation in vitro ( 17). Nm23-H1–mediated suppression of in vitro motility, colonization, and invasion has also been observed for other cell lines, including the aggressive breast carcinoma cell line MDA-MB-231 and colon and prostate carcinoma cell lines ( 2, 4, 7, 10, 12, 18).
How Nm23-H1 suppresses metastasis remains an area of intense research. Most biochemical analyses of Nm23-H1 focused on its in vitro inhibition of motility as an experimental end point because the capacity to migrate is a hallmark of metastatic spread. Although transfection of wild-type Nm23-H1 inhibited MDA-MB-435 cell motility, substitution mutants P96S and S120G did not ( 16). The histidine protein kinase activity of Nm23-H1 was coordinately reduced in the P96S and S120G mutants, supporting the hypothesis that this activity contributed to motility suppression ( 16, 19). Nm23-H1 binding to the human PRUNE protein and stimulation of its phosphodiesterase activity also correlated with its capacity to suppress MDA-MB-435 cell motility ( 20). Other activities of Nm23-H1, such as its nucleoside diphosphate kinase activity and its putative DNA transactivation, did not correlate with the motility suppression phenotype ( 21, 22).
Given the complexity of metastasis, multiple pathways downstream of Nm23-H1 presumably mediate its phenotypic effects. Here, we report a microarray analysis of gene expression changes in MDA-MB-435 cells expressing either wild-type or mutant Nm23-H1 compared with control vector. Nine genes encoding for either secreted factors or cell surface receptors were specifically down-regulated by wild-type Nm23-H1 and significantly correlated with Nm23-H1 gene expression in breast tumor cohort expression data. Of these genes, only reexpression of the LPA receptor EDG2 restored motility to Nm23-H1–suppressed tumor cells.
Materials and Methods
Cell lines and cell culture techniques. MDA-MB-435–derived cell lines C-100, H1-177, S-22, and I-205 ( 3, 22); and parental MDA-MB-435, MDA-MB-231, and Hs578T were maintained in DMEM (Invitrogen) supplemented with 10% fetal bovine serum (FBS; Invitrogen). Human breast cancer lines T47D, SKBr3, BT474, and MCF7, and the human lung fibroblast line MRC5, were maintained as described by American Type Culture Collection (ATCC). Human lung fibroblasts NHLF were maintained as described by Cambrex. Cells were grown in an atmosphere of 5% CO2 at 37°C to ∼85% confluency before harvesting for experiments.
Microarray analysis. Total RNA was extracted from cells using TRIzol reagent (Invitrogen) and chloroform and purified according to the RNeasy midiprep spin kit protocol (Qiagen). Oligo microarray chips were generated from the 34,580 longmer probe set Human Genome Oligo Set Version 3.0 (Qiagen). Dried oligos, resuspended in 3× SSC solution, were spotted to epoxy-coated slides. Slides were cleaned by vigorous shaking in a 0.5% SDS solution for 2 min, incubated in a 50°C waterbath for 20 min, and dried by centrifugation. Protocols for cDNA labeling, hybridization, and scanning are available through the National Human Genome Research Institute microarray core. 6 Three replicate hybridizations of C-100 v H1-177 samples were done and four replicate experiments each were done for C-100 v S-22 and C-100 v I-205 samples. Microarray data can be obtained via the Gene Expression Omnibus repository, accession no. GSE7549. 7 Raw intensities for Cy3 and Cy5 channels were normalized and data were filtered by eliminating array elements with a quality score below 0.5 and a total intensity of <100. Median Cy3 and Cy5 intensities for each element were determined from the replicate experiments. The log10 of the ratio of the medians for each element was calculated and a Student's t test was done to determine statistical confidence. Microarray elements with P < 0.05 and log10 ratios ≥0.23 were deemed differentially expressed compared with C-100 background.
Semiquantitative and quantitative reverse transcription-PCR. Semiquantitative and quantitative reverse transcription-PCR (qRT-PCR) analysis was done using total RNA isolated from C-100, H1-177, S-22, and I-205 cells. Reverse transcription was done according to protocol with SuperScript III Reverse Transcriptase (Invitrogen). Semiquantitative PCR conditions were optimized for gene oligo pairs (Supplementary Table S1) and analyzed by gel electrophoresis and ethidium bromide staining. qRT-PCR was done with an iQ5 Multicolor Real-time PCR Detection System (Bio-Rad) using standard techniques. The relative gene expression value of each gene was calculated by the standard curve of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) relative quantitation with C-100 cDNA.
Establishment of stable cell lines. The coding sequences of Nm23-H1, Nm23-H1-P96S, and Nm23-H1-S120G, previously cloned into pCDNA3.1, were subcloned into pCDNA3.1_Zeocin (Invitrogen) and sequence verified. MDA-MB-435 cells were selected for stable transfection of pCDNA3.1_Zeo, Nm23-H1_Zeo, P96S_Zeo, and S120G_Zeo, and MDA-MB-231 cells were selected for stable transfection of pCDNA3.1_Zeocin and Nm23-H1_Zeocin using standard procedures for Effectene (Qiagen). Single clones were selected in medium containing Zeocin (Invitrogen) and screened for Nm23-H1expression.
Immunoblotting. Lysates were prepared as described ( 23) and protein concentration was determined by a BCA Protein Assay (Pierce). Equivalent amounts of total protein were separated by SDS-PAGE and transferred to Hybond C-Extra nitrocellulose membrane (Amersham Biosciences). Membranes were then probed with anti–Nm23-H1/H2 (Chemicon), anti–Nm23-H1 (BD Biosciences Pharmingen), anti-tubulin (Oncogene), anti-MET (Cell Signaling), anti-EDG2 (Abcam), anti-SMO (Abcam), anti-PTN (Oncogene), anti-CTGF (R&D Systems), anti-L1CAM (Abcam), anti-MMP2 (Novus Biologicals), anti-CD44 (Abcam), anti-SPP1 (Novus Biologicals), anti-RSK1 (Chemicon), anti-histone H2B (Cell Signaling), anti-EDG4 (Abgent), anti-hemagglutinin (Covance), and anti–FLAG-horseradish peroxidase (Sigma). Relative protein expression was determined by spot densitometry.
Statistical correlation to breast tumor cohorts microarray data. Microarray expression data from the van't veer et al. ( 24) breast tumor cohort and the Perou et al. ( 25) cohort is publicly available. Tumor samples were ranked by the log10 of the normalized ratio for the Nm23-H1 gene element and cohorts were cut in half into tumors with “high” Nm23-H1 expression and “low” Nm23-H1 expression. To determine significance of the expression of a given gene to Nm23-H1 gene expression, a Student's t test and a Wilcoxon rank-sum test was done on the normalized ratio or the log10 of the normalized ratio of expression for its representative microarray element for the given gene in the low and high Nm23-H1 tumor sets. Scatterplots were constructed and trend lines were estimated in Microsoft Excel.
Gene silencing. Nm23-H1 expression was diminished in SKBr3 cells by transfection with pGENEClip containing either NME1-1 and NME1-2 short hairpin RNA (shRNA) constructs against Nm23-H1 coding sequence (SuperArray Bioscience) with standard Effectene procedures. Five days posttransfection, cells were collected and lysed.
For EDG2 silencing, MDA-MB-435 and MDA-MB-231 cells were transfected with small interfering RNA designed against the coding sequence of EDG2, available from Dharmacon, using HiPerfect (Qiagen). Transfection efficiency was monitored by siGLO, and siCONTROL, a nontargeting small interfering RNA (siRNA) pool, served as a negative control.
Immunohistochemistry. Liver tissues were obtained from ASV/Nm23-M1+/+ and ASV/Nm23-M1−/− mice as described in Boissan et al. ( 14) and embedded in paraffin blocks. Blocks were sliced into 4-μm sections and mounted to silanized glass slides. Sections were prepared and stained by immunohistochemistry as previously described ( 26). The following primary antibodies were used: anti–Nm23-H1 ( 14), anti-EDG2 (Novus Biologicals), anti-MET (R&D Systems), anti–NCAM-L1 (Santa Cruz Biotechnologies), antipleiotrophin (R&D Systems), and anti-Smo (Abcam). Stained sections were examined microscopically at ×400 magnification in consultation with a pathologist.
Cloning and expression of Nm23-H1–regulated genes. Seven genes were cloned into the mammalian expression vector pCDNA3.1 (Invitrogen) by PCR amplification with primers containing restriction enzyme adapter sequences, standard restriction enzyme digest, gel extraction, ligation, and transformation procedures (see Supplementary Table S2 for oligo sequences). EDG2, PTN, and L1CAM were amplified from clones available from ATCC and MMP2 was amplified from a clone that was a gift from L. Matrisian (Cancer Biology, Vanderbilt University Medical Center, Nashville, TN). FZD1, SMO, and NETO2 were amplified from a human placental cDNA library (Stratagene) and were simultaneously tagged using oligos containing linker epitopes. The CTGF and MET expression constructs were previously described ( 27, 28) and the EDG4 expression construct is available through ATCC. Constructs were transfected into H1-177 cells using standard Effectene (Qiagen) protocols.
Boyden chamber assays. Cell migration assays were done essentially as described ( 16). Lower wells contained DMEM plus 0.1% bovine serum albumin (BSA) with or without 0.1% FBS, 1% FBS, and 5 μmol/L LPA (Sigma). Cells (1 × 105) were added to each upper well in DMEM plus 0.1% BSA and incubated in a humidified chamber at 37°C, 5% CO2. Cells were fixed and stained with Diff-Quik solutions (Dade-Behring). Cells that had migrated to the undersurface of the membrane were examined microscopically at ×100 magnification. Each condition was assayed in triplicate wells and representative areas of each well were counted to determine the number of cells that had migrated. Statistical differences were calculated by a Student's t test and each experiment was repeated at least twice.
Matrigel invasion assay. NHLF and MRC5 cells were grown to 70% confluency, washed, and serum starved in DMEM containing 1 mg/mL BSA. Conditioned medium was collected and applied as a chemoattractant. C-100 and H1-177 transfected cells were grown to 85% confluency and 4 × 105 cells were applied to BD-Biocoat Matrigel-coated invasion chambers (BD Biosciences). Assays were incubated for 40 h at 37°C, 5% CO2. Cells that invaded through the Matrigel to the other side of the 8-μm porous membrane were fixed and stained with Diff-Quik solutions (Dade-Behring). The membranes were subsequently attached to a glass slide, examined microscopically at ×100 magnification. Each cell condition was assayed in triplicate chambers and the entire membrane was counted for invading cells. Statistical significance was calculated by a Student's t test and each experiment was done twice.
Genes differentially regulated in wild-type versus mutant Nm23-H1–expressing lines. The H1-177 cell line expressed 9-fold higher levels of Nm23-H1 and exhibited reduced motility in vitro and metastasis in vivo compared with vector transfectant, C-100 ( 3). Two stable MDA-MB-435–derived cell lines expressing mutant Nm23-H1 proteins, P96S and S120G (termed S-22 and I-205, respectively), exhibited motility levels not significantly different from vector transfectants ( 22). Immunoblot analysis of these cell lines showed that the Nm23-H1 transfectants retained higher levels of Nm23-H1 protein relative to the C-100 line ( Fig. 1A ). In chemotactic assays, stable transfection of wild-type Nm23-H1 reduced cell migration by ∼30-fold and 14-fold to 0.1% and 1% FBS, respectively (P = 0.002 and P = 0.03, respectively). Cell migration of the S-22 and I-205 cells was not statistically different from the control vector–transfected cell line at either concentration of FBS ( Fig. 1B). These data confirmed the correlation of Nm23-H1 mutational status and inhibition of in vitro motility for the cell passages used herein.
A microarray gene expression analysis of the H1-177 and mutant S-22 and I-205 cells compared with C-100 cells was done. A significant difference in expression between the wild-type and at least one mutant was determined for 70 genes by a Student's t test (Supplementary Table S3). These differentially expressed genes either alone or in combination may contribute to Nm23-H1–mediated motility suppression.
Validation of microarray analysis. To validate the microarray results, semiquantitative and quantitative RT-PCR was done on a subset of the Nm23-H1–regulated genes. Of 23 genes tested, 19 showed identical trends in all four stable cell lines when compared with the microarray data ( Fig. 1C; Supplementary Fig. S1 and data not shown). These genes include MET, CTGF, EDG2, FZD1, L1CAM, PTN, SMO, SPP1, and MRGX genes and histone genes. Expression of EDG2 was nearly undetectable in the H1-177 line by these methods, but was expressed at levels comparable with C-100 in the S-22 and I-205 mutant lines. Similarly, MET, L1CAM, PTN, and NETO2 expression was reduced in the H1-177 line, but not in S-22 and I-205 when compared with control transfectant, C-100. The four genes that failed to reproduce observed microarray trends, MAP2, NCAM, CCL20, and Nm23-H2, all exhibited divergent expression patterns in the S-22 cell line only.
The corresponding protein levels for 11 Nm23-H1–regulated genes was determined by immunoblot analysis using lysates from the same stable cell lines ( Fig. 1D). The trends in protein levels in all four cell lines precisely matched the trends observed in the microarray analysis for nine differentially expressed genes: MET, EDG2, L1CAM, SMO, CTGF, PTN, MMP2, CD44, and SPP1. In contrast to the microarray and RT-PCR analysis, RSK1 and Histone H2B protein levels in the S-22 and I-205 lysates were comparable with that observed in the H1-177 lysates. Notably, EDG2 does not migrate as a single, clean band, presumably due to posttranslational modifications. Three antibodies against EDG2 have yielded similar immunoblot results (data not shown).
To refute the possibility that the observed trends in protein levels were the result of clonal variation and not Nm23-H1 expression, an independent set of MDA-MB-435 transfectant cell lines was established. The level of Nm23-H1 expression and the migratory capacity of these lines paralleled that observed for the original clonal lines (Supplementary Fig. S2 and data not shown). Identical protein expression patterns were observed for MET, L1CAM, and EDG2 in these newly established lines (Supplementary Fig. S2). Although SMO and PTN levels were lower in the Nm23-H1_Zeo line compared with vector control as expected, their relative levels in the P96S_Zeo and S120G_Zeo lines were different from that observed in the original S-22 and I-205 lines.
Nm23-H1–regulated genes with relevance to breast cancer. To refine the set of genes linked to Nm23-H1 motility inhibition to genes with putative clinical relevance, the expression of these genes was correlated to Nm23-H1 gene expression in two published breast tumor cohorts: the van't Veer et al. ( 24) and the Perou et al. ( 25) data sets. The van't Veer cohort contains 120 breast tumors and the Perou cohort consists of 60 tumors. Each study measured relative gene expression of bulk tumor tissue by microarrays consisting of ∼24,500 elements and 9,200 elements, respectively.
To determine statistical significance between Nm23-H1 and downstream genes, cohorts were divided into two groups: high-expressing and low-expressing Nm23-H1 tumors, shown pictorially in Fig. 2A for the van't Veer cohort. The fold ratio of expression of all 70 Nm23-H1–regulated genes that were differentially expressed in at least one of the mutant cell lines were assessed for significance between the high- and low-expressing Nm23-H1 groups with a Student's t test and a nonparametric rank-sum analysis (Supplementary Table S4; Fig. 2B–D).
Expression of 14 Nm23-H1–regulated genes significantly correlated with Nm23-H1 expression in the van't Veer cohort and nine genes correlated with Nm23-H1 in the Perou cohort (Supplementary Table S4). Only three genes overlap between the cohorts, primarily because a number of genes are not represented on the Perou microarray. Six genes exhibited a significant relationship in one cohort, but no correlation in the other. These discrepancies may be explained by differences in the patient cohorts as the van't Veer cohort represents a younger population (median age of 44 versus 64 for the Perou cohort) or differences in stromal tissue contamination in the collected tumors. Regardless, this analysis showed that the observed patterns of gene expression corresponding to Nm23-H1 motility suppression can be found in clinical breast tumors and refine the list of Nm23-H1–regulated genes to 20 genes with putative clinical relevance. These genes include MET, EDG2, L1CAM, SMO, and PTN, which all negatively correlate with Nm23-H1 expression.
Correlation of Nm23-H1–regulated genes to metastatic potential and Nm23-H1 expression in a panel of breast carcinoma cell lines. The gene expression patterns relative to Nm23-H1 were next shown to be germane to a breast cancer cell line model. The expression of Nm23-H1, MET, EDG2, L1CAM, SMO, and PTN was assessed by immunoblot analysis in a panel of human breast cancer cell lines with variable metastatic capacities. As shown in Fig. 3A , the nonmetastatic breast carcinoma cell lines T47D, MCF-7, BT474, and SKBr3 have high levels of the Nm23-H1 protein, comparable with levels observed in H1-177 cells, whereas the metastatic lines MDA-MB-435, Hs578T, and MDA-MB-231 have low levels of this metastasis suppressor. MET, EDG2, SMO, and PTN exhibited an inverse correlation with Nm23-H1 and thus a positive correlation with metastatic capacity in all lines tested, in contrast to L1CAM.
Next, Nm23-H1 was overexpressed in the metastatic MDA-MB-231 human breast carcinoma cell line to determine its effect on Nm23-H1–regulated gene expression. Lysates were prepared from two vector-control cell lines (3 and 6) and three Nm23-H1–expressing lines (22, 28, and 38) and immunoblot analysis was done. MET, EDG2, L1CAM, SMO, and PTN levels were all diminished in the Nm23-H1–overexpressing clones relative to vector control ( Fig. 3B). Thus, the Nm23-H1–related variations in gene expression observed in MDA-MB-435 cells seem to be reproducible in a human breast carcinoma cell line.
Alternatively, it was hypothesized that silencing Nm23-H1 expression would enhance expression of EDG2 and MET in an independent cell line expressing high endogenous levels of Nm23-H1 protein. SKBr3 cells were transiently transfected with two Nm23-H1 shRNA constructs or an empty vector control and the expression of Nm23-H1, EDG2, and MET determined by immunoblot. Transfection of Nm23-H1 shRNA 1 and 2 resulted, respectively, in a 2.2- and 2.4-fold reduction in Nm23-H1 protein in SKBr3 cells relative to transfection of the vector control ( Fig. 3C). Proportional increases in EDG2 and MET were observed when Nm23-H1 levels were reduced.
Enhanced levels of EDG2, MET, L1CAM, and PTN in Nm23-M1−/− mouse HCC. Nm23-M1 is the murine homologue of Nm23-H1. An Nm23-M1 knockout mouse has recently been characterized in the context of metastatic disease. Nm23-M1−/− mice induced to form HCC with transgenic expression of simian virus large T antigen (ASV) have a much higher incidence of pulmonary metastasis compared with Nm23-M1+/+ mice similarly induced ( 14). Genes identified as Nm23-H1 regulated might contribute to the metastatic spread observed in the knockout mouse. To test this hypothesis, the relative protein levels of a number of genes down-regulated by Nm23-H1 were determined in ASV/Nm23-M1+/+ and ASV/Nm23-M1−/− mice by immunohistochemistry.
Regions of hepatocellular transformation in ASV/Nm23-M1−/− and ASV/Nm23-M1+/+ liver were identified in H&E-stained sections, in consultation with a pathologist. Induction with ASV resulted in transformation of much of the liver tissue of both the −/− and +/+ mice. As expected, Nm23-M1 staining was more intense in HCC from ASV/Nm23-M1+/+ mice compared with null mice ( Fig. 4 ). Side-by-side comparison of HCC regions from the null and wild-type mice revealed that EDG2, MET, L1CAM, and PTN staining was consistently higher in the ASV/Nm23-M1−/− tumor cells compared with ASV/Nm23-M1+/+ HCC and micrographs of two such regions from both the wild-type and null mice are represented in Fig. 4. Very little SMO staining was detected in the liver sections of mice with either genotype. In experiments not shown, EDG2 and MET protein levels were further validated in HCC sections from six ASV/Nm23-M1−/− and six ASV/Nm23-M1+/+ mice.
EDG2 restores in vitro cell motility to Nm23-H1–overexpressing cancer cells. To determine the functional effect that differential expression of these genes might have on Nm23-H1–mediated motility suppression, genes down-regulated by Nm23-H1 were cloned into a mammalian expression vector and their capacity to restore cell motility to wild-type Nm23-H1–overexpressing H1-177 cells was assayed in a Boyden chamber. MET, FZD1, SMO, EDG2, MMP2, L1CAM, NETO2, PTN, and CTGF constructs were transiently transfected into H1-177 cells and their expression was monitored by immunoblot analysis ( Fig. 5A ). Figure 5B summarizes the results of these transfections on H1-177 cell motility assayed in response to 1% FBS over 16 h. Most transfected genes exhibited no motility enhancing activity in H1-177 cells. However, exogenous overexpression of EDG2 stimulated a 60-fold increase in motility in response to serum (P = 0.0008). This increase in cell number cannot be attributed to enhanced proliferation as comparable increases were observed when motility was assayed over 4 h (data not shown). The only other gene that significantly enhanced the motility of the H1-177 cells was MET, the hepatocyte growth factor (HGF) receptor gene, which stimulated motility only 2-fold to serum (P = 0.008) and 4-fold to HGF (data not shown).
EDG2 is a G-protein coupled receptor that mediates LPA signaling. When pure LPA was used as a chemoattractant in the Boyden chamber motility assay, migration of EDG2-transfected H1-177 cells was not enhanced beyond that observed for serum, presumably because LPA is a major component of serum ( Fig. 5C). EDG2 also stimulated an increase in motility of H1-177 cells in the absence of a chemoattractant (BSA; Fig. 5B–C). This background migration may be attributed to the cell secretion of autotaxin, which converts fatty acids in BSA into the lipid mediator LPA, as fatty acid free BSA dramatically reduced this background motility (data not shown). Interestingly, overexpression of EDG2 in H1-177 cells also stimulated motility to HGF and heregulin by 8.6- and 6.8-fold above vector control transfectants (P = 0.0003 and 0.021, respectively) in the presence of fatty acid–free BSA. Although EDG2 transfection induced cell migration, it did not restore motility to the extent observed for C-100 cells ( Fig. 5C). This incomplete restoration could represent an artifact of transient transfection.
A major site of breast cancer metastasis is the lung. To better simulate in vivo conditions for metastasis, we assessed tumor cell invasion in response to lung fibroblast-conditioned medium. Conditioned medium from MRC5 and NHLF cells was used as a chemoattractant to determine the capacity of EDG2 to prompt invasion through Matrigel. H1-177 cells transfected with EDG2 promoted invasion of the H1-177 cells, 12-fold over control transfectants (P = 0.00007; Fig. 5D).
EDG2 is essential for the motility of metastatic cancer cells. We next sought to determine if EDG2 expression was essential for tumor cell migration. Expression of EDG2 was inhibited by transfection of MDA-MB-435 and MDA-MB-231 cells by a siRNA against EDG2 coding sequence (siEDG2), and its reduced expression was monitored by immunoblot analysis ( Fig. 6A ). Compared with transfection of a control siRNA, siCONTROL, siEDG2 reduced the motility of MDA-MB-435 and MDA-MB-231 cells by ∼70% (P = 0.002 and P = 0.0002, respectively; Fig. 6B). siEDG2 did not completely diminish cell motility, presumably because transient transfection of siRNA does not completely abolish EDG2 expression in all cells. Of note, transient transfection of Nm23-H1 in these cells resulted in a comparable reduction in migration.
A second possibility is that another LPA receptor compensated for reduced EDG2 levels. Four human LPA receptors have been identified, which include homologues EDG2, EDG4, and EDG7 and nonhomologous GPR23 ( 29– 32). EDG4 has been reported to have the highest affinity for LPA ( 33). EDG4 mRNA expression was not significantly different in the H1-177 cells compared with C-100 cells and hybridization was below the threshold of detection for the oligo corresponding to the 3′ region of the EDG7 gene for both channels. Immunoblot analysis of lysates from both sets of MDA-MB-435–derived clonal cell lines confirmed similar expression levels for EDG4 between Nm23-H1– and vector-transfected cell lines (data not shown).
As a complementary approach, the capacity of EDG4 expression to restore motility to H1-177 cells was determined by a Boyden chamber assay. Although EDG4 enhanced the motility of these cells ∼10-fold (P = 0.004), it was not nearly as effective as EDG2 transfection, despite similar levels of overexpression (a 4-fold increase for EDG2 versus a 3.5-fold increase in EDG4; Fig. 6C–D). Therefore, restoration of a motility phenotype to H1-177 cells seems to specifically require the LPA receptor EDG2.
Nm23-H1 has been implicated in multiple aspects of cancer spread—it inhibits tumor cell motility, invasion, and anchorage-independent growth, as well as prompting tumor cell differentiation. Given its complicated array of phenotypes, the molecular mechanism of Nm23-H1–mediated metastasis suppression is likely complex, involving the coordinated regulation of multiple genes and pathways. As a step toward comprehensively delineating the genes involved in this process, we identified a set of genes dependent on functional, motility suppressive Nm23-H1 in an aggressive cancer cell line, MDA-MB-435.
Much of the characterization of Nm23-H1 activity and the analysis of the molecular basis of breast cancer metastasis have been done in the MDA-MB-435 cell line. This cell line is controversial as some reports have indicated that it may have been confused with the M14 melanoma cell line at some point ( 34, 35). However, MDA-MB-435 cells have been shown to produce milk lipid droplets upon induction of differentiation, and Nm23-H1 overexpression itself has been shown to stimulate the hallmarks of breast differentiation in these cells ( 17, 36). Moreover, this line consistently forms spontaneous metastases when implanted into the mouse mammary fat pad. Regardless, it is still a valuable and well-characterized model of metastatic progression.
Gene expression analysis in this model system revealed that the expression of nearly 200 genes is dependent on Nm23-H1. The list of Nm23-H1–regulated genes was refined to those with functional relevance to metastasis by two methods. First, we capitalized on the finding that the P96S and S120G mutants of Nm23-H1 do not suppress motility and cross-compared the expression profiles of wild-type and mutant Nm23-H1–expressing cell lines. Second, this subset of genes was correlated with Nm23-H1 expression in two published microarray data sets of breast tumor cohorts. Although this analysis is somewhat limited given that not much gene overlap was observed between the cohorts (Supplementary Table S4) and Nm23-H1 gene expression only showed a trend with patient survival (P = 0.06, data not shown), it sufficiently narrowed the list of genes to nine that had an inverse correlation to Nm23-H1. Strong correlations between the expression of a number these genes and Nm23-H1 expression were exhibited in vivo in the Nm23-M1 knockout mouse model of HCC and in vitro in a panel of breast carcinoma cell lines exhibiting differential metastatic propensities.
The expression of the LPA receptor EDG2 inversely correlated with Nm23-H1 expression in all of these models and more importantly, overexpression of this receptor restored a motile and invasive phenotype to Nm23-H1–suppressed tumor cells. EDG2 transfection stimulated an astonishing 60-fold increase in the motility of these cells. To date, only two other genes have significantly abrogated the motility-suppressive phenotype of Nm23-H1-PRUNE and EBNA3C, and these genes only enhanced motility 2- and 3-fold, respectively ( 20, 37). Both are Nm23-H1 binding proteins and may sequester Nm23-H1 rather than participate in its mechanistic pathway. The data reported herein identify for the first time a protein that overcomes the suppression of motility at points downstream of Nm23-H1.
EDG2 may largely explain the motile phenotype of metastatic MDA-MB-435 and MDA-MB-231 cells as silencing this gene severely inhibits the motility of these cells. In addition, LPA and serum, which contains high LPA titers, are the most potent chemoattractants of MDA-MB-435 cells, enhancing motility 50- to 200-fold above background compared with the 6- to 7-fold increases observed for other growth factors, which include HGF, heregulin, IGF, and PDGF ( 15). Interestingly, EDG2 expression is capable of stimulating motility in Nm23-H1–suppressed cells, albeit to a lesser extent, in LPA-limiting conditions, such as in the absence of chemoattractant or with HGF and heregulin. EDG2 may stimulate this observed motility either by facilitating signaling through other receptors or by responding to cellular secretions of autotaxin, the phospholipase involved in LPA biosynthesis ( 38, 39).
The functional contribution of EDG2 to in vivo metastasis will be determined in future experiments, but several lines of evidence suggest that LPA and EDG2 play an important role in the promotion of metastasis. LPA levels are significantly increased in malignant effusions and exogenous expression of autotaxin promotes metastasis in nude mice ( 40– 42). Similarly, in an orthotopic ovarian tumor model, LPA supplied through microosmotic pumps enhanced the incidence of metastases ( 43, 44). Up-regulation of autotaxin and LPA receptor genes has been noted in some cancers, including breast cancer ( 39, 45, 46). More poignantly, EDG2 overexpression prompted bone metastasis of the breast cancer cell line MDA-BO2 in xenograft experiments ( 47, 48). Down-regulation of this receptor by the metastasis suppressor Nm23-H1 is critical for motility suppression and may therefore be important for the suppression of in vivo metastasis.
Grant support: Intramural Research Program of the National Cancer Institute, Center for Cancer Research, NIH.
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.
We thank T. Clair, S. Davis, P.S. Chen, M.L. Kuo, S. Kanda, and I. Lascu.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received March 13, 2007.
- Revision received May 2, 2007.
- Accepted May 29, 2007.
- ©2007 American Association for Cancer Research.