Purpose: S100 proteins are implicated in metastasis development in several cancers. In this study, we analyzed the prognostic role of mRNA levels of all S100 proteins in early stage non–small cell lung cancer (NSCLC) patients as well as the pathogenetic of S100A2 in the development of metastasis in NSCLC.
Experimental Design: Microarray data from a large NSCLC patient cohort was analyzed for the prognostic role of S100 proteins for survival in surgically resected NSCLC. Metastatic potential of the S100A2 gene was analyzed in vitro and in a lung cancer mouse model in vivo. Overexpression and RNAi approaches were used for analysis of the biological functions of S100A2.
Results: High mRNA expression levels of several S100 proteins and especially S100A2 were associated with poor survival in surgically resected NSCLC patients. Upon stable transfection into NSCLC cell lines, S100A2 did not alter proliferation. However, S100A2 enhanced transwell migration as well as transendothelial migration in vitro. NOD/SCID mice injected s.c. with NSCLC cells overexpressing S100A2 developed significantly more distant metastasis (64%) than mice with control vector transfected tumor cells (17%; P < 0.05). When mice with S100A2 expressing tumors were treated i.v. with shRNA against S100A2, these mice developed significantly fewer lung metastasis than mice treated with control shRNA (P = 0.021).
Conclusions: These findings identify S100A2 as a strong metastasis inducer in vivo. S100A2 might be a potential biomarker as well as a novel therapeutic target in NSCLC metastasis.
Distant metastasis is a major contributor to cancer related death. Here, we identify S100A2 as a prometastatic gene in patients with non–small cell lung cancer (NSCLC): High S100A2 expression is closely linked to poor survival in surgically resected NSCLC. Expression of S100A2 in NSCLC cell lines induces a highly metastatic phenotype in a murine xenograft model. Repression of S100A2 reverts the metastatic phenotype in vivo indicating that S100A2 is primarily responsible for metastasis development. These findings suggest that S100A2 is an important contributor to metastasis development in NSCLC. It might be possible to use S100A2 as a biomarker for metastasis risk and as a therapeutic target to specifically inhibit metastasis development.
Lung cancer is the leading cause of cancer death. Less than 20% of patients with non–small cell lung cancer (NSCLC) survive 5 years (1, 2). The development of metastasis after initial surgery is the main reason for cancer related death in early stage NSCLC. The process of metastasis is still not completely resolved, although several metastasis models exist in vivo (3). Several steps are required for the metastatic process: Initially, tumor cells must invade into the surrounding tumor tissue, enter either the lymphatics or the bloodstream and extravasate into a new tissue to settle and grow at this new site. Currently, these steps can partially be reproduced in vitro using assays for invasion and migration (4, 5).
The S100 protein family is a multigenic group of nonubiquitous cytoplasmic EF-hand Ca2+-binding proteins comprising 21 known human members each coded by a separate gene. At least 16 of these genes cluster to chromosome 1q21, known as the epidermal differentiation complex (6). They are differentially expressed in a wide variety of cell types and have been reported to be involved in the regulation of inflammatory responses (7) and in cellular processes such as cell-cycle progression and differentiation (8). Expression of several S100 proteins seems to be altered in different types of cancers including lung adenocarcinomas (9–11). In addition, some S100 proteins have been shown to be associated with metastasis (12–18). In 1989, Ebralidze et al. (14) indicated that the S100 protein S100A4 (mts1) is involved in regulating the metastatic behavior of tumor cells. Recently, we reported S100P and S100A2 overexpression in a small set of metastasizing tumors (12).
An additional indication for their involvement in inflammatory and neoplastic disorders is that most S100 genes are found near a break-point region on human chromosome 1q21, which, if affected, is responsible for a number of genetic abnormalities related to autoimmune pathologies or cancer (6). Although the function of S100 proteins in cancer cells is still unknown, the specific expression patterns of these proteins are a valuable prognostic tool (19).
Concerning S100A2, conflicting results have been published. Feng et al. (20) suggested S100A2 to be a putative tumor suppressor at an early stage of human lung carcinogenesis. Other groups described an association between the down-regulation of S100A2 and the development of melanoma (16) and other malignant cells (21).
In this study, we analyzed the prognostic role of mRNA levels of all S100 proteins in early stage NSCLC patients. S100A2 emerged as a gene that is closely associated with poor survival in surgically resected NSCLC. S100A2 also induced a metastatic phenotype in vitro and in vivo indicating a prominent role for S100A2 in the metastatic process in NSCLC.
Materials and Methods
Patient data and statistical analysis of patient data. The mRNA expression data for S100 proteins as analyzed by Affymetrix microarrays was extracted from the National Center for Biotechnology Information Gene Expression Omnibus (22). A total of 196 patients (male, 123; female, 72; missing, 1) were analyzed. Adenocarcinoma was diagnosed in 138 patients and squamous cell carcinoma in 58 patients. Additional patient information data were kindly provided by Drs. A. Potti and J.R. Nevins (Duke University, Durham, NC). Most patients were diagnosed with stage I disease (stage I, 120 patient; stage II, 12 patients; stage 3A, 11 patients; stage 3B, 8 patients; stage 4, 22 patients; missing information, 23 patients). Additional information on the patient cohort can be found in the original publication (23). Expression data for S100 proteins [S100A11, S100A10, S100A13 (two probesets), S100A8 (two probesets), S100A4, S100A9, S100A2, S100P, S100A1, S100A12, S100A7, S100A3, S100A5, S100G, S100A11, S100B, S100A6, S100A14, S100PBP] were extracted from the normalized data based on the respective Affymetrix gene identifiers and loaded into SPSS 14.0 (SPSS, Inc.). Expression data of all the probesets was analyzed for overall survival differences using a stepwise (Wald) Cox regression analysis that was stratified for patients' stage at the time of diagnosis. Kaplan Meier plots at the 75th percentile of expression were calculated and evaluated using the log-rank test. S100PBP is not a real S100 protein. Its exclusion from the Cox regression analysis did not alter the results (data not shown).
Cell culture. HTB56 and HTB58 lung adenocarcinoma cells were cultured at 37°C, high humidity, and 5% CO2 in Modified Eagle's medium (Invitrogen). The medium was supplemented with 10% FCS, 1% streptomycin and penicillin, 1% glutamine, 1% sodium pyruvate, and 1% nonessential amino acid. HMEC-1 cells were cultured at 37°C, high humidity, and 3% CO2 using MCDB 131 medium (Invitrogen) with 1% streptomycin and penicillin, 5% glutamine, 10% FCS Gold, and 50 μg/mL gentamicin.
Cloning and transfection. The coding sequence of S100A2 (NM_005978) was cloned into the expression vector pcDNA3.1(+), fused at the NH2 terminus with enhanced green fluorescent protein (EGFP). Stable transfection of the cloned construct into HTB56 (Calu6) was done with Lipofectamine reagent (Invitrogen) and selected with Neomycin (G418/Sigma). Bulk cultures were used for all experiments to avoid clone-specific effects. Expression levels were determined by quantitative real-time reverse transcription-PCR, Western blotting, and Flow cytometry. Silencing of gene expression was achieved using shRNA technology. Oligonucleotides targeting S100A2 (sense, GATGAGAACAGTGACCAGCAG; antisense, CTGCTGGTCACTGTTCTCATC; loop, TTGATATCCG) and the scrambled control (sense, AGATCCGTATAGTGTACCTTA; antisense, TAAGGTACACTATACGGATCT; loop, TTGATATCCG) were synthesized by Invitrogen, were annealed, and were cloned into the shRNA expression vector pRNAT-H1.1/Neo (GenScript). HTB58 cells were transfected and selected as described above. Expression levels were verified by quantitative real-time reverse transcription-PCR.
Gene expression analysis. Analysis of gene expression was done using quantitative reverse transcription-PCR as described (24). The primer and probe (FAM and TAMRA labeled) used were as follows: S100A2 forward (5′- CTGGGTCTGTCTCTGCCACC), reverse (5′-GCAGGAGTACTTGTGGAAGGTAGTG), and probe (5′- FAM-TGCCACAGATCCATGATGTGCAGTTCT-TAMRA). The mRNA expression levels were calculated with regard to the internal standard of glyceraldehyde-3-phosphate dehydrogenase as described (25).
Western blot analysis. Proteins were detected using the following antibodies: S100A2 (Abcam), GFP (Santa Cruz), Stat5 (Santa Cruz), and Actin (Sigma-Aldrich) as first antibodies, Goat anti-mouse and Goat anti-rabbit (both from Dianova) as secondary antibodies. Western blot analysis was carried out as described (24).
3[H]-thymidine incorporation. A total of 2 × 104 stably transfected HTB56 cells were incubated overnight at 37°C and 5% CO2 with 200 μL medium (Modified Eagle's medium with 0.5% FCS) in a 96-well plate. The next day 0.037 Mbq (1 μCi) 3[H]-thymidine was added to each well, and cells were incubated for additional 8 h. Cells were harvested onto glass fiber filters, and β-emission of bound DNA was analyzed by the Wallac 1450 microβ liquid scintillation counter. The data represent the mean of three independent experiments done in hexaplets.
Migration assay. A total of 5 × 105 cells (in 100 μL Modified Eagle's medium with 5% FCS) were seeded into the upper part of a Transwell chamber (transwell filter inserts in 6.5 mm diameter with a pore size of 5 μm; Corning, Inc.), which was 30 min precoated with 50 μg fibronectin. In the lower part of the chamber, 600 μL Modified Eagle's medium with 20% FCS (a serum gradient was used as chemoattractant) was added and the assay was done for 16 h at 37°C and 5% CO2 before migrated cells were analyzed by flow cytometry. For the migration assay using transendothelial cells, transwell filters were precoated with 2.2 × 105 HMEC-1 cells for 48 h at 37°C and 3% CO2, forming a confluent transendothelial cell layer (5). NSCLC cells (5 × 105) were stained for 5 min with 5 μmol/L SNARF (1 carboxylic acid, acetate, succinimidyl ester; Molecular Probes) to avoid counting fragments of endothelial cells and added to the upper part of the transwell chamber coated with transendothelial cells. After 16 h of migration, cells were harvested by trypsinization and analyzed (only SNARF-positive cells) by flow cytometry. All assays were done in triplicate and independently done for thrice.
Metastasis experiments in vivo. For all mouse experiments, we used 8- to 10-wk-old NOD.CB17-Prkdc<scid>/J [NOD/severe combined immunodeficient (SCID)] mice obtained from Charles River.
To analyze metastasis development following primary tumor removal, 2.5 × 106 stably transfected NSCLC cells (supplemented in 200 μL PBS) were injected s.c. into the right flank. After 4 wk of tumor growth, primary tumors were surgically removed and the tumor weight (grams) was determined. Mice were followed for 8 additional wk (12 wk after initial tumor injection). At this time, mice were sacrificed, and tumor weight and metastasis development was determined. Metastasis development was evaluated by counting individual metastatic nodules. For histologic analyses, the lungs were fixed in 4% paraformaldehyde. To analyze shRNA effects on metastasis development upon i.v. tumor cell injection, NOD/SCID mice were irradiated with a single dose of 3.5 Gy from a cobalt-60 unit 1 d before transplantation. A total of 2 × 106 stable transfected cells (EGFP-control cells, dissolved in 200 μL PBS) were injected i.v. into the tail vein. One week after transplantation, 20 μg shRNA (dissolved in 250 μL PBS) were i.v. injected. Treatment with shRNA was repeated four times every fourth day. Subsequent to the treatment, mice were followed for an additional 5 wk before metastasis development was analyzed. In a few cases, mice died: 1 mouse transplanted with the S100A2-transfected HTB 56 cells died during surgery, 2 other mice died the following day after they were i.v. injected with overexpressing cells for S100A2. Therefore, the number of the mice differs by one or two. None of the mice died in the week after shRNA injection. In all experiments, treatment groups were randomized to prevent cage effects.
Statistical analysis. All experimental data are shown as mean plus SD if not indicated otherwise. The mean values of two groups were compared by t test. In the few cases where the variances differed significantly, the nonparametric Mann-Whitney U test was used and stated together with the P value. When more than two groups were compared, one way ANOVA was used. A P value of <0.05 was considered significant.
Subsequent to surgery, death in early stage NSCLC can in most cases be attributed to metastasis development (25). Several S100 proteins have been implicated in metastasis, but a general assessment has been missing. Therefore, we analyzed a large microarray expression data set of surgically resected NSCLC with respect to the association between S100 mRNA levels and lung cancer related death in early NSCLC.
To identify survival associated S100 proteins in this data set, we did an exploratory analysis with a stage adjusted Cox regression model with all S100 proteins whose mRNA expression was analyzed by the microarray data. In this analysis, high expression levels of S100A2 (P = 0.001), S100P (P = 0.009), S100A12 (P = 0.002), and S100B (P = 0.026) were independently associated with poor prognosis. The association between gene expression and survival for these genes was subsequently assessed in Kaplan-Meier plots. Patient samples with high S100A2 expression (>75th percentile) showed a significantly inferior overall survival compared with low-level expressing patients (Fig. 1A). Differences for S100P and S100A2 were marginally significant, whereas no survival differences were found for different levels of S100B expression (Fig. 1A, bottom left). Overall, 41.9% (62 of 148 patients) of the patients with low-level expression of S100A2 died, whereas 60.4% of patients with high expression died (P = 0.031; Fig. 1B). The phylogenetic tree of human and murine S100 proteins is shown in Fig. 1C. Interestingly, S100A2 is related to S100A4, a known metastasis inducing gene (13–15).
Because these findings indicated a potentially specific role for S100A2 in metastasis, we further analyzed the metastatic properties of S100A2. First, we expressed S100A2 in the HTB56 adenocarcinoma cell line, which is only weakly metastatic. Establishing a stable cell line of HTB56 expressing S100A2 was achieved by cloning this gene, fused with EGFP, into the expression vector pcDNA3.1(+). As control cells, we used HTB56 cells expressing EGFP alone. After transfection into HTB56, cells were selected with Neomycin and sorted by flow cytometry. Bulk cultures were used to avoid clone specific effects. S100A2 expression was verified by Western blotting, real-time PCR, and flow cytometry. Both cell lines expressing S100A2-EGFP and EGFP alone showed a corresponding band on the protein level using a GFP antibody (Fig. 2A). Overexpression of S100A2 was confirmed both on the protein and on the mRNA level (Fig. 2A and B). Flow cytometry analysis displayed at least 90% EGFP positivity of the bulk culture (Fig. 2C). In relation to the control cell line, no alteration regarding growth properties or proliferative capacity was recognized in the cell line overexpressing S100A2 (data not shown; Fig. 2D). To analyze migration, we used a two-chamber assay (Transwell chamber) where cells migrate through a microporous filter from the upper into the lower chamber. To mimic the in vivo situation, we also did assays in which the cells had to migrate through an endothelial monolayer grown on the microporous filter.
In both sets of migration experiments, S100A2 significantly enhanced migration, i.e., transwell migration as well as transwell migration through an endothelial layer toward a serum gradient, compared with the control cells expressing only EGFP (P < 0.005, t test; Fig. 2E).
Because in vitro experiments cannot recapitulate the complexity of metastasis, we used a recently established metastasis model (27) that involved surgical removal after tumor formation resembling early stage NSCLC treatment (Fig. 3A).
After transplantation of 2.5 × 106 stable NSCLC cells (HTB56 cells expressing EGFP or S100A2-EGFP) into the right flank of NOD/SCID mice, tumor growth was monitored for 4 weeks. Afterwards, the tumor nodule was surgically removed. In contrast to our previous findings regarding S100P, neoangiogenesis was not significantly enhanced in tumors overexpressing S100A2 (data not shown).
Subsequent to the surgery, the mice were monitored for 8 additional weeks. At this, time lungs and other organs were analyzed for the development of distant metastasis. Distant metastasis developed in 17% (2 of 12 mice developed lung metastasis) of control gene–transfected HTB56 cells. In contrast, metastasis formation was >3-fold increased in mice initially transplanted with the S100A2-transfected HTB56 cells with 64% (7 of 11) of the mice with S100A2-expressing tumors developing metastasis (Fig. 3B and C).
Next, we analyzed whether shRNA directed against S100A2 could be used to reduce metastasis development in S100A2-expressing tumors. First, shRNAs against S100A2 or the scrambled version (control) were tested in vitro. HTB58 cells were used in these experiments because these expressed relatively high levels of S100A2. HTB58 cells stably transfected with shRNA against S100A2 (cloned into the siRNA expression vector pRNAT-H1.1/Neo) were found to suppress S100A2 at the mRNA level (Fig. 4A). In addition, endogenous S100A2 protein expression was strongly reduced in cells transfected with shRNA against S100A2 compared with the control cells (Fig. 4B and C).
After the identification of active and specific shRNA against S100A2, a total of 2 × 106 HTB56 cells were injected i.v. into the tail vein of a NOD/SCID mouse. One week after transplantation, 20 μg shRNA (dissolved in 250 μL PBS) was i.v. injected. The treatment with shRNA was repeated every fourth day for a total of four injections. In parallel experiments, 10 mice each were treated with shRNA against S100A2 or the scrambled control. Six weeks after i.v. injection of HTB56-EGFP cells, the lungs of the mice treated with scrambled shRNA were all infiltrated with metastasis. In all mice (10 of 10) lungs were penetrated with at least 3 and up to 11 metastases. In contrast, 25% of mice (2 of 8) treated with shRNA against S100A2 did not develop metastases (Fig. 4D). Furthermore, the numbers of metastasis was significantly decreased in these mice, treated with shRNA against S100A2 (P = 0.021, t test; Fig. 4E). The treatment with shRNA against S100A2 reduced the metastasis counts in the lungs by ∼70%. These results suggested that S100A2 was relevant for metastasis development in this mouse model.
Metastasis is the most common cause for tumor-related death. In lung cancer as well as in other common cancers such as breast and colon, patient prognosis ultimately depends on the occurrence or absence of metastasis (28). Metastasis is a complex phenomenon that is still not well-understood. Preliminary studies show that metastasis is not only dependent on an individual gene; rather it seems that complex networks are involved in this process. Recently, our group and others identified metastasis-related genes using mircroarray expression analysis and other genome wide techniques (12, 23, 29, 30). We could show that S100 proteins, in particular S100A2 or S100P, were associated with a significantly increased risk for NSCLC patients to develop distant metastasis (12). It is interesting to note that several S100 proteins are overexpressed in human tumors (9, 16, 31–37) and that this expression correlates with tumor progression (15, 38–40) and metastasis (12–14, 17). In the current study, we used a large data set of NSCLC patients to identify metastasis associated S100 proteins. We also analyzed the similarities between metastasis-associated and other S100 proteins. Phylogenetic tree algorithms do not cluster metastatic versus nonmetastatic S100 proteins. Rather, specific nonconserved domains might be relevant.
Our previous microarray analyses implicated S100A2 as a metastasis-associated gene (12). Other studies had reported conflicting results indicative of either tumor-suppressive or tumor-promoting functions of S100A2 (20, 41, 42). The close association between S100A2 expression and metastasis/death in a large, independent data set from the group of A. Potti and JR Nevins (23) now provides additional evidence for the metastasis-associated phenotype of S100A2. These analyses are in accordance with recently published data (11). The authors observed increased expression of S100A2 in early-stage tumors of patients with NSCLC. Many lung adenocarcinomas expressed S100A2 (41). Correlation of S100A2 with progression and poor prognosis was also reported for pancreatic cancer (42).
Based on these findings, we analyzed the involvement of S100A2 in prometastatic functions in vitro and in vivo. Indeed, S100A2 enhanced migration and transendothelial migration in vitro. Importantly, in a xenograft mouse model, distant metastasis was found to be >3-fold increased in mice with S100A2-expressing tumors.
Often, the process of metastasis is described to be associated with the recruitment of new blood vessels, angiogenesis (43–46). No significant increase in blood vessel formation in tumor samples from our mice positive for metastasis could be observed. Indeed, the vessel area per visual field was increased, but compared with our recent results with S100P (27), no significant association between metastasis and angiogenesis was observed. Nonetheless, S100A2 was functionally important in metastasis development because in vivo treatment with shRNA against S100A2 inhibited metastasis development.
Taken together, our studies provide evidence that S100A2 plays an important role in metastasis development and, thus, could be a valuable drug target. Its close association with patient survival in independent data sets further highlights its potential importance and its usability as a biomarker. Because S100A2 did not induce angiogenesis, the mode of action of S100A2 might differ, at least in part, from S100P.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: Non–small cell lung cancer research in our laboratory is funded by the Deutsche Forschungsgemeinschaft and the Deutsche Krebshilfe. This project was partially supported by the Wilhelm Sander-Stiftung.
We thank Dr. J.R. Nevins and Dr. A. Potti from the Duke Institute for Genome Sciences and Policy, Duke University, Durham, North Carolina for providing us clinical data from NSCLC patients that were analyzed in microarray experiments.