Correspondence to: Wangjun Liao, Department of Oncology, Nanfang Hospital, Southern Medical University, Guangzhou 510515, People's Republic of China, Tel: +86–2062787731, Fax: +86–2062787731, E-mail: firstname.lastname@example.org
Metastasis-associated in colon cancer-1 (MACC1) is a newly identified oncogene, and little is known about its role in gastric cancer (GC). Our study was performed to investigate whether MACC1 influences the prognosis of GC patients and to explore the potential mechanisms involved. MACC1 expression was verified to be higher in GC tissues than in adjacent nontumorous tissues by Western blotting. A retrospective analysis of 361 GC patients (Stages I–IV) revealed that higher MACC1 expression was associated with more advanced disease, more frequent postoperative recurrence, more metastases and a higher mortality rate. The disease-free survival of Stage I–III patients and overall survival of Stage-IV patients were significantly worse when their tumors showed high MACC1 expression. To investigate the underlying mechanisms, MACC1 overexpression and downregulation were established in two GC cell lines (BGC-823 and MKN-28 cells). MACC1 overexpression significantly accelerated tumor growth and facilitated metastasis in athymic mice. MACC1 also promoted the proliferation, migration and invasion of both GC cell lines. Moreover, gastric MACC1 mRNA expression levels were significantly correlated with markers of the epithelial-to-mesenchymal transition (EMT) in patients with GC. MACC1 overexpression upregulated mesenchymal–epithelial transition factor and induced changes to markers of EMT, whereas silencing of MACC1 reversed all these changes. These findings provide some novel insights into the role of MACC1, a gene that contributes to a poor prognosis of GC by promoting tumor cell proliferation and invasion as well as the EMT.
Metastasis-associated in colon cancer-1 (MACC1), a novel identified oncogene, has been reported to promote tumor proliferation and invasion mediated via hepatocyte growth factor (HGF)/mesenchymal–epithelial transition factor (c-Met) signaling in colorectal cancer.[1-3] However, its role in gastric cancer (GC) is largely unknown. Recently, a clinical study of GC patients showed that MACC1 was frequently expressed in tumors with peritoneal dissemination, which aroused our curiosity to investigate whether MACC1 is associated with the progression of GC, and if so, to determine the underlying mechanisms.
It is generally believed that the epithelial-to-mesenchymal transition (EMT) is an essential process for tumor cell invasion. During the invasion process, cancer cells must free themselves from cellular junctions and the surrounding matrix, and must breach these barriers by changing from an epithelial to a mesenchymal phenotype. HGF/c-Met signaling has been proved to be a downstream target of MACC1 in colon cancer. Moreover, HGF/c-Met is known to be a tumorigenic and prometastatic factor for GC.[6, 7] Deregulation of HGF/c-Met has been reported to induce the EMT.[7-10] However, it remains unclear whether MACC1 can modulate the EMT process.
In our study, we aimed to examine the role of MACC1 in GC and the potential mechanisms involved by a retrospective analysis of 361 GC patients categorized from Stage I to Stage IV, and by carrying out animal and in vitro experiments to clarify the influence of MACC1 on GC proliferation and invasion and its effect on the EMT.
Material and Methods
Patients and tissue specimens
Our study was approved by the Nanfang Hospital Ethics Review Board. Tissue samples for diagnostic purposes were obtained with the consent of each patient. All these patients had been histologically diagnosed as GC at Nanfang Hospital (Guangzhou, Guangdong, China) from 2000 to 2011, and tumor staging was defined according to of the AJCC Cancer Staging Manual (the 7th edition, 2010). Among them, 264 Stage I–III patients received radical resection (43, 86 and 135 for Stages I, II and III, respectively), and 97 Stage-IV patients (the metastasis-affected distant organs include peritoneum, liver, distant lymph node, transverse colon, ovary or oviduct, pancreas, bone, lung, brain, intestine and inferior vena cava) underwent palliative surgery and/or chemotherapy. The median follow-up time was 22.4 months (from 0.5 to 80.0 months).
Immunohistochemical staining was carried out with the Dako Envision System (Dako, Glostrup, Denmark) and a rabbit polyclonal antibody for MACC1 (Abnova, Taipei, China) as described previously. Target protein expression level in tumor tissues was scored by a semi-quantitative method as described elsewhere with modifications. Briefly, for MACC1 staining intensity, sections were scored as 0 (negative), 1 (weak), 2 (medium) or 3 (intense), whereas the staining extent was scored according to the area percentages: 0 (0%), 1 (1–25%), 2 (26–50%), 3 (51%–75%) or 4 (76–100%). The products of the staining intensity and extent scores were the final staining scores (0–12) for MACC1 expression. Tumors of final staining score ≥3 were considered to be positive expression because 95% of normal gastric tissues expressed low level of MACC1 with an IHC score of <3 in our preliminary study. We further defined 3–7 as low expression and 8–12 as high expression to perform qualitative analysis.
Western blot analysis
Tissues and cells were lysed and subjected to Western blotting as described previously. Primary antibodies were as follows: rabbit anti-MACC1 (Abnova); mouse anti-E-cadherin, mouse anti-α-Catenin, mouse anti-fibronectin and mouse anti-vimentin (BD Biosciences, Franklin Lakes, NJ); rabbit anti-CD44 (Genetex, Hsinchu, China); mouse anti-matrix metalloproteinase 2 (MMP2), rabbit anti-MMP9 and rabbit anti-β-actin (Santa Cruz Biotechnology, Santa Cruz, CA). Using the enhanced chemiluminescence method with a Western blotting detection system (Kodak Digital Science, Rochester, NY), the immunoreactive bands were visualized and quantified by Image software QuantityOne v4.6.2.
Human gastric adenocarcinoma cell lines of poorly differentiated BGC-823 and well-differentiated MKN-28 were obtained from Foleibao Biotechnology Development (Shanghai, China). Cells were cultured in complete medium (Roswell Park Memorial Institute 1640 medium [Invitrogen, Life Technologies, Carlsbad, CA] with 10% fetal bovine serum [Thermo Scientific HyClone, South Logan, UT]) and incubated under 5% CO2 at 37°C. Cells were collected in logarithmic growth phase for all experiments as described in the following sections.
Establishment of stably transfected cell lines
For MACC1 overexpression, ectopic MACC1 coding sequence was amplified by polymerase chain reaction (PCR) (primer sequences in Supporting Information Table S1), and cloned into the pBaBb-puromycin plasmid. For MACC1 silencing, sequences of short hairpin RNA targeting MACC1 (shMACC1) and scramble were cloned into the pSUPER-retro-puromycin plasmid (sequences in Supporting Information Table S2). Cell lines were transfected with aforementioned constructed plasmids combined with PIK vector or blank pBaBb-vector. Stably transfected cell lines were selected with 0.5 µg/mL (a minimum lethal dose) puromycin at 48 hr after infection. By this selection criterion, MACC1 was markedly increased in the MACC1 overexpression group and greatly inhibited in MACC1 silencing group in the transfected GC cells (Supporting Information Fig. S1).
To evaluate whether MACC1 affects tumor growth in vivo, BGC823 cells of MACC1 and vector control or shMACC1 and scramble control (∼1 × 107) were subcutaneously and bilaterally inoculated into 6-week-old male athymic mice (five mice per cell line) on the flank regions of legs. Tumor nodules were monitored every 3 days by caliper measurements of the length and width of the tumors. Tumor volumes were calculated according to the formula: Volume = width × length × (width + length)/2. On day 18, the mice were euthanized and tumors were harvested. To evaluate whether MACC1 affects tumor metastasis in vivo, caudal intravenous injection (∼2 × 106 cells/mouse, n = 6) of BGC823 cells with different MACC1 expression levels was performed in athymic mice. On day 40 after caudal intravenous injection, the mice were euthanized and dissected. Metastasis nodules in different organs of mice were counted and confirmed by hemotoxylin and eosin staining.
Soft agar assay for colony formation
When trypsinized and suspended in 2 mL of complete medium with 0.3% agar, cells were plated onto a bottom layer with 1% agar (Sigma, St. Louis, MO) in complete medium. For 4 weeks of culturing, cells were observed and photographed under a microscope (Nikon, Tokyo, Japan).
Cell viability analysis
After trypsinization, cells were seeded and cultured on 96-well plates at an initial density of 0.2 × 104/well. Capacity of cell proliferation was measured by methyl thiazolyl tetrazolium (MTT) assay on days 1, 3, 5 and 7. For this, 0.02 mL of MTT solution (5 mg/mL in phosphate-buffered solution [PBS]) was added to each well, and incubated for 4 hr at 37°C. Thereafter, the medium was replaced by 0.15 mL of dimethyl sulfoxide for 10-min incubation. The optical density at 492 nm was measured by Microplate spectrophotometer (Thermo Scientific, Franklin, MA). All experiments were performed in triplicate.
The cultured cells were harvested in PBS and fixed in cold ethanol (70%) for overnight. After washed in PBS, cells were permeabilized with 0.1% Triton-X-100 and 100 U/mL RNAase in PBS for 30 min at 37°C in the absence of light, and then cells were stained with 50 µg/mL of propidium iodide (Sigma) for 30 min. The cell-cycle phases were analyzed by flow cytometry system (Beckman Coulter, Indianapolis, IN) at an excitation wavelength of 488 nm and an emission wavelength of 525 nm.
BrdU incorporation assay
Cells were plated on coverslips (Fisher Scientific, Pittsburgh, PA). After 48-hr serum starvation, cells were incubated with 10 µmol/L BrdU for 1 hr and stained with anti-BrdUrd antibody (Upstate, Temecula, CA) according to the manufacturer's instruction. The total nuclei were labeled by DAPI (BD Biosciences). The experiment was repeated three times independently, and the stained cells were counted under the Olympus BX51 fluorescence microscope (Olympus Optical, Tokyo, Japan).
Wound closure assay
The trypsinized cells were seeded on six-well culture plates at a density of 5 × 104 cells per/well, and cultured in complete medium. Upon confluence, cells were scratched with a sterile 200-µL pipette tip. Wound closure was observed at 0, 12 and 24 hr under an inverted microscope.
Transwell invasion assay
Transwell invasion assay was performed using Boyden's chambers. Cells were planted in the upper chamber consisting of 8-mm membrane filter inserts (1 × 105/mL, 200 µL/well) coated with Matrigel (BD Biosciences). The chemoattractant in lower chamber was supplemented with medium containing 10% fetal bovine serum. After 37°C incubation for 24 hr, cells invaded through the coated membrane to the lower surface were fixed with 4% paraformaldehyde and stained with hematoxylin. The results were photographed under microscope and the invading cells were counted offline.
3D morphogenesis assay
Cells were trypsinized and seeded at a density of 2 × 104 cells per well in 500 µL of medium on the growth factor-reduced solidified Matrigel (BD Biosciences). After 2-week culturing, the cells were photographed under a microscope.
Quantitative Real-Time PCR
Primer sequences of MACC1, E-cadherin, α-cadherin, fibronectin, MMP2, MMP9, vimentin, CD44 and β-actin are summarized in Supporting Information Table S3. Total RNAs were extracted from cultured cells or human GC samples using a Trizol kit (Bio-Rad Laboratories, Hercules, CA) according to the manufacturer's recommended protocol. cDNAs were synthesized by using total RNAs and MMLV-RT reverse transcriptase (ProSpec, East Brunswick, NJ). The reaction mixture for real-time PCR was prepared by following the manufacturer's recommended protocols.
Results are reported as the mean ± SEM or median. The difference significances were calculated using the χ2 test or rank test for categorical variables and ANOVA or Student's T-test for continuous variables. Survival and recurrence rates were analyzed according to the Kaplan–Meier method and examined by the log-rank test. Hazard ratio (HR) was determined using Cox model. In addition, the least squares method was used to analyze linear of the selected variables. All p-values of <0.05 were considered statistically significant. Data were analyzed using the SPSS16.0 software package.
MACC1 expression is associated with a poor prognosis of GC
To confirm whether MACC1 was expressed in GC, we performed Western blotting of surgical samples. Expression of MACC1 was found to be significantly higher in tumor tissues than in the adjacent nontumorous gastric tissues (Fig. 1a). Then, MACC1 protein expression in normal gastric tissue and GC tissue was assessed by immunohistochemical staining. In normal gastric tissue, MACC1 was seldom detected, whereas its expression was markedly upregulated in GC tissue. Representative examples of staining with high and low magnifications are shown in Figure 1b and Supporting Information Figure S2. By scoring the immunostaining of 361 GC samples, we noted that MACC1 positivity was more frequent in GC patients with more advanced stages (p = 0.001), recurrence (p < 0.001), metastasis (p = 0.002), and those who were dead at the completion of follow-up (p = 0.017) (Fig. 1c). Stage-IV patients scored significantly higher than either Stage I or Stage II–III patients. The staining scores were also higher for patients with recurrence, metastasis and death (Fig. 1d). Clinical follow-up showed that patients with higher MACC1 expression in GC tissues had a worse prognosis. Clinical features are summarized in Supporting Information Table S4. According to the multivariate analysis using the Cox model, MACC1 expression (p = 0.001, HR = 1.999) and tumor stage (p < 0.001, HR = 2.060) were independent factors predicting the risk of recurrence after radical resection in Stage I–III patients, whereas MACC1 expression (p = 0.048, HR = 2.217) and metastasis (p < 0.001, HR = 4.718) were independent predictors of mortality in Stage-IV patients (Supporting Information Table S5). Accordingly, these parameters were assessed by Kaplan–Meier analysis (Figs. 1e–1h). In Stages I–III, the median time to recurrence was 30.0 months. The median time to recurrence for patients with MACC1-positive tumors was 24.0 months, and 60% of the MACC1-negative patients were relapse-free at the end of follow-up (Fig. 1e). Among Stage-IV patients, the survival rate of those with MACC1 expression was significantly lower than that of patients without MACC1 expression, and the median survival time was 10.2 versus 24.0 months, respectively (Fig. 1g). To determine whether MACC1 is an oncogene or a tumor suppressor gene for GC, animal in vivo experiments were carried out.
MACC1 promotes GC cell growth and metastasis in athymic mice
Using the BGC-823 GC cell line, colonies of ectopic-MACC1 and shMACC1 and their respective controls were prepared. For tumor growth assessment, GC cells were injected subcutaneously into athymic mice, and all successfully formed tumors on the flanks near back legs (Fig. 2a). On day 6, the growth of ectopic-MACC1 clones began to accelerate, whereas shMACC1 clones were markedly suppressed. Up to day 18, the average volume of the ectopic-MACC1 tumors was markedly larger than that of the vector controls, and shMACC1 injected mice displayed significantly smaller tumors than the scramble controls (Fig. 2b). For the metastasis evaluation, we injected BGC823 cells via caudal vena into the body of athymic mice. On day 40, postmortem examination showed that metastatic nodules frequently appeared in lungs, and far less in liver (Figs. 2c and 2d). Only in the ectopic-MACC1group, the metastasis could occur in both organs (Fig. 2d). The pulmonary metastasis was validated by histological examination (Supporting Information Fig. S3), and the number of pulmonary metastatic nodules in the ectopic-MACC1group was significantly more than in other three groups. Metastasis was significantly inhibited in shMACC1 group manifested by significant less metastatic nodules in the lungs (Figs. 2c and 2d).
MACC1 promotes the proliferation of GC cell lines
In both cell lines, proliferation was found to be promoted by ectopic-MACC1 and inhibited by shMACC1, as evidenced by the number and size of the colonies (Figs. 3a and 3b). As demonstrated by the MTT assay, the number of viable cells was significantly increased by MACC1 upregulation, but was decreased by shMACC1 (Fig. 3c). Regarding the effect on the cell cycle, MACC1 accelerated progression into the S phase (DNA synthesis), whereas shMACC1 caused cells to exhibit growth arrest in G1 phase (Fig. 3d). These findings were further confirmed by the BrdU incorporation assay, showing that newly synthesized DNA was significantly increased by MACC1 overexpression (Fig. 3e and Supporting Information Fig. S4).
MACC1 enhances the migration and invasion of GC cell lines
Our clinical and in vivo experimental findings indicate that MACC1 expression is associated with GC metastasis, but the mechanism remains unknown. Thus, cell migration experiments were carried out using both cell lines to clarify this issue. In the 2D wound closure migration assay, ectopic MACC1-transfected cells almost closed the wound after 24 hr of culture, whereas the vector-transfected cells left a wide gap. In contrast, migration by shMACC1-transfected cells was significantly impaired, as evidenced by the fact that wound recovery was greatly suppressed in comparison with the scramble control (Fig. 4a). The effect of MACC1 on cell invasion was also assessed. When cultured in 3D Matrigel, ectopic MACC1-infected cells displayed an aggressive morphology with protrusion of longer podosomes. In contrast, shMACC1-infected cells became spheroids with few protrusions (Fig. 4b). Invasive activity was further verified by a Transwell assay. The number of invading cells that migrated through the Matrigel was significantly greater for ectopic-MACC1 cells versus vector controls, and was markedly decreased for shMACC1 cells versus scramble cells (Fig. 4c). Next, the potential mechanisms involved were examined.
MACC1 stimulates c-Met expression and is involved in the EMT
We examined mRNA expression levels of MACC1, c-Met and the markers of EMT in GC samples obtained from 22 recent patients who received gastric resection in our hospital. We found that MACC1 was positively correlated with c-Met (r = 0.921, p < 0.001) (Fig. 5a), and the regulation effect of MACC1 on c-Met was confirmed in GC cell lines (BGC-823 and MKN-28) (Fig. 5b). In addition, significant correlations were found between MACC1 and E-cadherin (r = 0.539, p = 0.010), fibronectin (r = 0.447, p = 0.037) and vimentin (r = 0.713, p < 0.001), respectively (Fig. 5c). Moreover, E-cadherin (r = 0.456, p = 0.033), fibronectin (r = 0.552, p = 0.008) and vimentin (r = 0.814, p < 0.001) were also significantly correlated with c-Met (Fig. 5d).
As significant correlation is not necessary to indicate a causative relationship, we then further verified whether the EMT was induced by MACC1 in GC cells lines. In both cell lines, changes of mRNA expression (Supporting Information Fig. S5) and protein expression (Figs. 6a and 6b) were consistently observed. By knocking down the MACC1 gene, expression of fibronectin, MMP2, MMP9, vimentin and CD44 was significantly downregulated, whereas E-cadherin and α-catenin were upregulated. In contrast, overexpression of MACC1 significantly increased expression of fibronectin, MMP2, MMP9, vimentin and CD44, whereas inhibiting E-cadherin and α-catenin.
Up till now, analysis of a small population of GC patients has demonstrated that the MACC1 gene is associated with peritoneal dissemination. However, little is known about its role in GC. In our study, the oncological significance of this gene for GC was investigated and it was demonstrated that (i) MACC1 expression was significantly upregulated in GC, (ii) MACC1 upregulation was associated with a shorter survival time and a higher incidence of recurrence and metastasis, (iii) MACC1 contributed to both proliferation and invasion of GC cells and (iv) MACC1 modulated the EMT.
MACC1 was first identified as a colon cancer oncogene that promoted metastasis and proliferation. Recently, MACC1 has also been found in other cancers. In hepatocellular carcinoma, MACC1 is thought to be a prognostic biomarker for survival. In lung adenocarcinoma, a higher MACC1 level has been shown to have a correlation with postoperative recurrence. These findings regarding the role of MACC1 in different malignant tumors support the clinical results of our study, implying that MACC1 may ubiquitously promote carcinogenesis, at least for tumors of the digestive system.
In our study, alteration of MACC1 expression provided further functional evidence, supporting the stimulatory effect of MACC1 on GC proliferation and invasiveness at the cellular level and in an animal model.
MACC1 was first identified as a regulator of HGF/c-Met signaling that stimulates c-Met expression, which was verified in the GC cell lines in our study. Deregulation of HGF/c-Met signaling triggers various malignant behaviors in cancer, one of which is the EMT.[7-10] The EMT is considered to be an initiating step for the cascade of tumor invasion driven by alterations of gene expression that control epithelial and mesenchymal plasticity. During the typical EMT, epithelial-like tumor cells lose polarity, acquire mesenchymal characteristics and become motile and invasive as single cells. Protrusion of podosomes is frequently found in EMT cancer cells and tumor metastasis is promoted. At this time, cancer cells display a more active podosome structure when MACC1 overexpression was induced. On the other hand, when MACC1 was silenced, podosomes were largely inhibited. Therefore, it was reasonable to speculate that modulation of GC metastasis by MACC1 is associated with the EMT process.
To determine whether these tumor characteristics initiated by MACC1 are associated with the EMT, we focused on hallmarks of the EMT phenotype. Our study showed that various EMT-associated proteins were modulated by MACC1 as the process required.
Intercellular adhesions are critical for the maintenance of epithelial-like morphology, whereas loss of these junctions allows aggregated cells to separate into single cells. E-cadherin is required for adherent junctions, whereas its deficiency on cytomembrane would result in the loss of cell polarity and abnormal differentiation, and finally facilitates the EMT.[17, 18] Moreover, membranous E-cadherin deficiency usually accompanied by dysfunctional E-cadherin upregulation in cytoplasm as a compensative manner, and this phenomenon can be post-transcriptional regulated. In our study, the result that E-cadherin mRNA was positively correlated with MACC1 and c-Met in GC samples likely reflects such compensation. As the linkage between the E-cadherin and the actin cytoskeleton, α-catenin maintains the epithelial tissue structure by zonula adherens and regulates Rho signaling, which modulates the transition between epithelial and mesenchymal phenotypes.[21, 22] We showed that both adhesive molecules of E-cadherin and α-catenin were downregulated by MACC1 in the cell-line models, implicating that MACC1 plays a causative role in epithelial characteristics weakening.
Extracellular matrix (ECM) remodeling is another process of the EMT. Fibronectin contributes to ECM remodeling and is upregulated during the EMT, inducing MMPs activation. MMPs cleave ECM components, modulate the niche and facilitate mesenchymal cell migration. These changes were also observed in our study where MACC1 overexpression induced elevation of fibronectin, MMP2 and MMP9 expression, whereas MACC1 inhibition suppressed these ECM-associated markers.
For cells to achieve mesenchymal features, vimentin, an intermediate filament, is required for remodeling of the cytoskeleton elongation of invadopodia. CD44 overexpression enables cancer cells to acquire stem cell properties of self-renewal and the formation of differentiated progeny,[18, 28] become active in niche remodeling, and develop resistance to chemotherapy and radiation. After upregulation of MACC1, we found that these two representative mesenchymal markers were consistently increased.
Although our study did not investigate whether these EMT markers are directly modulated by MACC1 or indirectly through other signals, such as HGF/c-Met, previous studies have conclusively shown that some of these MACC1 transcription targets regulate HGF/c-Met. For instance, when binding to integrins, fibronectin activates c-Met and its downstream signals, and thereby facilitates cancer cell invasion across tissue boundaries. CD44v6 is a CD44 isoform that is strictly required for c-Met activation by HGF in rat and human cancer cells. Therefore, the HGF/c-Met-induced EMT is enhanced by such feedback systems. Hence, it can be postulated that MACC1 regulates HGF/c-Met signaling via multiple pathways, and that because MACC1 initiates such feedback cycles, it may be a potential therapeutic target for inhibiting the EMT in cancer cells.
Based on our current cellular experiments, it is not difficult to suggest how the oncogene MACC1 affects the malignancy of GC, as originally shown in our clinical data. MACC1 enhances cancer cell invasiveness and facilitates niche remodeling, and hence cancer cells form micrometastases that are generally occult. Stem cell-like characteristics may be partially acquired via the EMT, which contributes to self-renewal and survival of immune clearance. In turn, the EMT is enhanced by CD44-HGF/c-Met and fibronectin-HGF/c-Met feedback loops. In addition, promotion of tumorigenicity enables cancer cells to withstand apoptotic and therapeutic stresses, as well as other antioncogenic compensations (Fig. 6c).
In conclusion, our study provided both clinical and experimental evidence that MACC1 acts as an oncogene in GC. MACC1 modulates cellular invasiveness via the EMT.