Tumor cell-derived angiopoietin-like protein ANGPTL2 is a critical driver of metastasis

Strategies to inhibit metastasis have been mainly unsuccessful in part due to insufficient mechanistic understanding. Here we report evidence of critical role for the angiopoietin-like protein ANGPTL2 in metastatic progression. In mice, Angptl2 has been implicated in inflammatory carcinogenesis but it has not been studied in human tumors. In lung cancer patients, elevated levels of ANGPTL2 expression in tumor cells within the primary tumor were associated with a reduction in the period of disease-free survival after surgical resection. Transcription factors NFATc, ATF2 and c-Jun upregulated in aggressive tumor cells promoted increased Angptl2 expression. Most notably, tumor-cell derived ANGPTL2 increased in vitro motility and invasion in an autocrine/paracrine manner, conferring an aggressive metastatic tumor phenotype. In xenograft mouse models, tumor-cell derived ANGPTL2 accelerated metastasis and shortened survival, whereas attenuating ANGPTL2 expression in tumor cells blunted metastasis and extended survival. Overall, our findings demonstrated that tumor-cell derived ANGPTL2 drives metastasis and provided an initial proof of concept for blockade of its action as a strategy to antagonize the metastatic process. In the present study, we investigated the role of ANGPTL2 in human tumor cells, and found that lung cancer patients showing high ANGPTL2 expression in cells within the primary tumor sites showed poor prognosis in terms of disease-free survival. Furthermore, we found that Angptl2 expression in tumor cells is induced by NFATc. Tumor cell-derived ANGPTL2 enhanced tumor cell motility and invasive capacity and increased tumor angiogenesis. Tumor cell-derived ANGPTL2 also accelerated metastasis and shortened survival periods in tumor cell-implanted mouse models. By contrast, decreasing ANGPTL2 levels in tumor cells attenuated metastasis and prolonged survival periods. Collectively, our findings provide strong evidence that tumor cell-derived ANGPTL2 worsens clinical prognosis and suggest that blocking ANGPTL2 could represent a novel therapeutic strategy to inhibit tumor metastasis.


Introduction
Cancer is a leading cause of death and accounts for 7.6 million deaths (approximately 13% of all deaths) worldwide (1). Although both diagnosis and therapeutic modalities used to treat cancer have remarkably improved, tumor metastasis still represents a major cause of cancer mortality (2,3). Therefore, identification of mechanisms underlying metastasis is essential to understand the pathophysiology of this lethal condition and identify novel therapeutic targets.
Inflammation plays key roles at various stages of tumor development, including initiation, growth, invasion, and metastasis (4). Recently, we found that angiopoietin-like protein 2 (ANGPTL2) increases inflammatory carcinogenesis in a chemically-induced skin squamous cell carcinoma (SCC) mouse model through enhanced susceptibility to "pre-neoplastic change" and "malignant conversion" (5). In addition, we also reported that ANGPTL2 expression in tumor cells is highly correlated with the frequency of tumor cell metastasis to distant organs and lymph nodes through increased tumor angiogenesis and tumor cell epithelial-to-mesenchymal transitions (EMT) (5). However, it has been obscure whether ANGPTL2 contributes to human cancer pathogenesis.
The nuclear factor of activated T-cell (NFATc) consists of five members (NFATc1-c4 and NFAT5). Among these factors, NFATc1-c4 function in tumor cell development and metastasis (6,7). For example, NFATc1 and NFATc3 contribute to the pathogenesis of melanoma and pancreatic cancer (8,9), NFATc2 plays in breast cancer cell migration and invasion (10), and NFATc4 promotes breast cancer cell growth (11). Research.
on August 13, 2017. © 2012 American Association for Cancer cancerres.aacrjournals.org Downloaded from Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.
In the present study, we investigated the role of ANGPTL2 in human tumor cells, and found that lung cancer patients showing high ANGPTL2 expression in cells within the primary tumor sites showed poor prognosis in terms of disease-free survival. Furthermore, we found that Angptl2 expression in tumor cells is induced by NFATc. Tumor cell-derived ANGPTL2 enhanced tumor cell motility and invasive capacity and increased tumor angiogenesis. Tumor cell-derived ANGPTL2 also accelerated metastasis and shortened survival periods in tumor cell-implanted mouse models. By contrast, decreasing ANGPTL2 levels in tumor cells attenuated metastasis and prolonged survival periods. Collectively, our findings provide strong evidence that tumor cell-derived ANGPTL2 worsens clinical prognosis and suggest that blocking ANGPTL2 could represent a novel therapeutic strategy to inhibit tumor metastasis. Research.
on August 13, 2017. © 2012 American Association for Cancer cancerres.aacrjournals.org Downloaded from Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Quantitation of ANGPTL2 protein by ELISA
ANGPTL2 concentrations in tissue lysates or in culture medium from tumor cells were estimated by an ANGPTL2 Assay kit (IBL), as described (12,13). For tissue lysates proteins were extracted from 2 mg of tumor or non-tumor tissue and dissolved in 10 ml lysis buffer (300 mM NaCl, 50 mM Tris-HCl (pH 7.5), 1% Triton X-100, and 1mM EDTA).

Cell lines and cell culture
The human lung cancer cell lines NCI-H460 and NCI-H460-LNM35 cells, as previously described (15)

Real-time quantitative RT-PCR
Real-time quantitative RT-PCR was performed as described (12).
Oligonucleotide primers are listed in Supplementary Table S1.

Cell invasion assay
Cell invasiveness was estimated using a 96-well BME Cell Invasion Assay kit (Trevigen) according to the manufacturer's protocol. Fluorescence was measured using a Fluoroskan Ascent fluorometer (Thermo Labsystems).  Cell migration was monitored by time-lapse microscopy using an Olympus IX-81 inverted microscope with a 20x objective lens. Phase contrast images were collected with CoolSNAP-HQ (Roper Scientific) at 3 min intervals. A series of time-lapse images was converted to video format using MetaMorph 6.1 software.

Time-lapse microscopy and FRET imaging
Cell motility was quantified using MetaMorph 6.1 software. FRET imaging was performed as described (13).

Statistics
The Kaplan-Meier log-rank test was used to analyze survival data of mice and patients using JMP7 software (SAS Institute). Data presented as means ± standard deviation (SD) or means ± standard error of mean (SEM) were analyzed using Student's t-test or analysis of variance (ANOVA). A P value of less than 0.05 was considered significant. Endo et al.

ANGPTL2 levels within primary tumor sites correlate with poor disease-free survival
To examine whether ANGPTL2 is expressed in human tumor tissues, we compared ANGPTL2 protein levels in lung tissues extracted from primary tumor sites in non-small-cell lung cancer (NSCLC) patients to those from peripheral non-tumor lung tissue using an enzyme-linked immunosorbent assay (ELISA).
ANGPTL2 protein levels in primary tumor sites were significantly greater than those seen in non-tumor lung tissue (Fig. 1A). Given the variation in ANGPTL2 expression in tumor tissue (Fig. 1A), we examined ANGPTL2 expression in lung tumor tissues by immunohistochemistry. Accordingly, we observed wide variation in the proportion of cells within a tumor that express ANGPTL2 (Fig.   1B, Supplementary Fig. S1). In situ hybridization analysis (12)

NFATc induces Angptl2 expression in lung cancer cells
To investigate tumor cell regulation of Angptl2 expression, we constructed luciferase reporter plasmids containing an Angptl2 regulatory region that we identified from the human NSCLC cell line NCI-H460 (H460) ( Supplementary   Fig. S4A). An F4 construct containing nucleotides -168 to +98 relative to the transcription start site of human Angptl2 showed high reporter activity in H460 cells, while activity of an F5 construct (containing -21 to +98) showed significantly decreased activity, indicating that F4 contains elements regulate Angptl2 expression in this context ( Supplementary Fig. S4A). In that region we identified potential binding sites for ATF/CREB, NF-κB, and NFAT (top in Fig.   2A). To investigate whether these factors affect Angptl2 expression, we transfected H460 cells with the F4 construct plus expression vectors encoding CREB, ATF2, ATF3, NF-κB, or a constitutively active form of NFATc (CA-NFATc). The CA-NFATc vector increased Angptl2 reporter activity relative to a pcDNA3.1 control plasmid, whereas the other factors had no effect ( Fig Endo et al.

ATF2 and c-Jun enhance NFATc-dependent Angptl2 expression in lung cancer cells
It has been reported that AP-1 components, c-Jun or c-Fos, form a stable heterodimer with ATF2, ATF3, or ATF4, and bind to the ATF/CREB site (18,19). NFATc and AP-1 heterodimers form a transcriptional complex that synergistically activates target genes (20)(21)(22). We found that reporter activity of F4 constructs containing a mutant ATF/CREB site was markedly decreased relative to wildtype F4 constructs ( Supplementary Fig. S4B), suggesting that ATF/CREB site is required for Angptl2 expression. We observed that CA-NFATc-dependent induction of Angptl2 reporter activity was also significantly suppressed in F4 constructs bearing a mutant compared to a wildtype ATF/CREB site (Fig. 2C). Furthermore, PMA/Ion treatment increased F4 reporter activity, an effect significantly decreased when we used F4 constructs harboring a mutant ATF/CREB site (Fig. 2D). Overall, it could be possible that NFATc forms a complex with AP-1 heterodimers that enhances NFATc-dependent induction of Angptl2 expression through the ATF/CREB site.
To investigate whether ATF family proteins may bind to ATF/CREB site on the NSCLC cell lines (23,24). c-Jun binding to the human Angptl2 promoter region was observed in PMA/Ion-untreated H460 cells, whereas that binding level was unchanged in PMA/Ion-treated cells (Supplementary Fig. S5A). Taken together, these observations demonstrate that NFATc plays an important role in Angptl2 induction and that an ATF2/c-Jun complex likely enhances NFATc-dependent Angptl2 induction through the ATF/CREB site.

NFATc2 knockdown decreases Angptl2 expression in tumor cells
The NCI-H460-LNM35 (LNM35) line was established by in vivo selection as a highly metastatic subline of the human large cell carcinoma of the lung, NCI-H460 (15). We found that that severe immuno-deficient Jak3-deficient NOD-SCID (NOJ) mice (25) subcutaneously injected with LNM35 cells showed shortened survival periods due to high frequency metastasis to lung and lymph nodes compared to mice injected with H460 cells (Supplementary Fig. S6).
Interestingly, Angptl2 mRNA is more abundant in cultured LNM35 cells compared to H460 cells (Fig. 3A). We found that NFATc1, NFATc2, and NFATc3 increased Angptl2 promoter activity in the H460 cells ( Supplementary   Fig. S5B and C). NFATc2 expression in LNM35 cells was significantly higher than that seen in H460 cells, while expression of NFATc1, NFATc3 and NFATc4 was equivalent in both lines (Fig. 3B). NFATc2 protein levels were also markedly increased in LNM35 cells, but ATF2 and c-Jun protein levels were equivalent (Fig. 3C). To examine whether NFATc2 is required for Angptl2 expression in LNM35 cells, we established two LNM35 cell lines each stably expressing a different miR RNAi expression vector designed to knockdown NFATc2 (LNM35/miNFATc2-1 and LNM35/miNFATc2-2) (Fig. 3D) Endo et al.
expression levels in both lines were significantly decreased compared with control cells and comparable to those seen in H460 cells (Fig. 3E). Invasive activity of LNM35/miNFATc2 cells was decreased compared with controls ( Supplementary Fig. S7). These observations suggest that NFATc2 is important for Angptl2 expression and acquisition of tumor invasivity.

Angptl2-expressing tumor cells contribute to increased tumor metastasis
As indicated in Fig.1 and Supplementary Fig. S2, ANGPTL2-positive tumor cells exhibit higher metastatic capacity than ANGPTL2-negative cells. We therefore generated two independent H460 lines constitutively expressing Angptl2 (H460/Angptl2-1 and -2) and an H460 line expressing a control vector (H460/Cont) (Fig. 4A). Although no differences were observed in in vitro growth among all three lines ( Supplementary Fig. S8A), in vitro invasive capacity of the two H460/Angptl2 lines was significantly greater than that seen in H460/Cont (Fig. 4B). We next analyzed tumor development after subcutaneous injection of either the three lines into mice. Despite equivalence in in vivo tumor growth observed among the two H460/Angptl2 lines and H460/Cont ( Supplementary   Fig. S8B and C), lung metastasis was more severe in mice bearing H460/Angptl2-1 tumors than in H460/Cont controls ( Fig. 4C and D). Tumor angiogenesis was also increased in mice bearing H460/Angptl2-1 compared to H460/Cont cells ( Supplementary Fig. S8D and E), a finding consistent with our recent report (5). Finally, the survival period of mice bearing H460/Angptl2 tumors was significantly shortened compared to H460/Cont mice (Fig. 4E).

Angptl2 increases tumor cell motility by activating Rac
Research.

Angptl2 knockdown suppresses tumor metastasis
Because LNM35 cells express Angptl2 abundantly and exhibit high metastatic capacity due to enhanced polarized morphology compared to H460 cells (Fig. 3,   Supplementary Fig. S6 and 9), we examined the effect of Angptl2 knockdown in Endo et al.
( Supplementary Fig. S10). We therefore generated two independent Angptl2 knockdown lines: LNM35/miAngptl2-b1 and -b2 ( Supplementary Fig. S11A). In vitro invasive ability was significantly decreased in both lines compared to LNM35/miLacZ cells, whereas no differences in in vitro cell growth were observed among knockdown and control lines ( Supplementary Fig. S11B and   C). In addition, both knockdown lines displayed less-polarized morphology and more uniform distribution of F-actin and paxillin compared with LNM35/miLacZ cells (Supplementary Fig. S12). Mice bearing both Angptl2 knockdown lines showed no significant difference in tumor growth but exhibited markedly decreased lung metastases compared to mice bearing LNM35/miLacZ cells (Supplementary Fig. S11D-G). Survival times were also extended in mice injected with knockdown versus control lines (Supplementary Fig. S11H). These results suggest that tumor cell-derived ANGPTL2 increases metastasis and that decreased ANGPTL2 expression in tumor cells can attenuate that effect.

Tumor cell-derived ANGPTL2 enhances lung metastasis in mice bearing breast cancer cells
To investigate whether tumor cell-derived ANGPTL2 enhances metastasis in other cancer types, we examined ANGPTL2 expression and function in the human breast cancer lines T47D, MDA-MB453 and MDA-MB231. Only MDA-MB231, which shows an aggressive metastatic phenotype (27), abundantly expressed and secreted ANGPTL2 (Fig. 6A and B). MDA-MB231 cells also expressed NFATc2, ATF2 and c-Jun (Fig. 6A) Endo et al. (Fig. 6C, Supplementary Fig. S13A and B). We established two independent MDA-MB231 lines expressing miR RNAi (miAngptl2-b) to knockdown Angptl2 (MB231/miAngptl2-b1 and -2) and a control LacZ RNAi line (MB231/miLacZ). In knockdown lines cellular Angptl2 mRNA levels and levels of ANGPTL2 protein in the culture medium were significantly decreased compared to those seen in controls (Supplementary Fig. S13C). All three lines showed similar in vitro growth ( Supplementary Fig. S13D), whereas in vitro invasive capacity was significantly decreased in knockdown compared to control cells ( Supplementary Fig. S13E).

MB231/miLacZ cells
Next, we performed in vivo xenograft tumor experiments by implanting various lines of MDA-MB231 cells into the mouse mammary fat pad. For imaging purposes, we induced a luciferase expression vector into MB231/miAngptl2-b1 and -2 and control MB231/miLacZ cells to create MB231/miAngptl2-b1/luc, MB231/miAngptl2-b2/luc, and MB231/miLacZ/luc, respectively. No difference in in vivo tumor growth was seen among the three lines (Fig. 6D, Supplementary   Fig. S13F). By contrast, lung metastasis was observed in mice bearing MB231/miLacZ/luc 5 weeks after tumor implantation, whereas fewer metastatic sites at the same time point was observed in mice bearing MB231/miAngptl2-b1/luc. To see equivalent lung metastasis after tumor implantation in mice bearing MB231/miAngptl2-b1/luc required eight weeks (Fig.   6D). Immunohistochemistry with an anti-CD44 antibody, which detects MDA-MB231 cells (27), showed statistically significant decreases in lung colonization of MB231/miAngptl2-b1 compared to MB231/miLacZ cells (Fig. 6E and F). Moreover, decreased tumor angiogenesis was observed in mice bearing Endo et al.

Discussion
Here, we show that tumor cells expressing ANGPTL2 exhibit high metastatic potential through acquisition of invasive and high cell motility phenotypes in an autocrine/paracrine manner. We also demonstrated that NFATc, ATF2 and c-Jun induce Angptl2 expression, providing a mechanism for tumor cell ANGPTL2 induction. These findings are consistent with previous reports showing that activation of ATF/CREB family proteins and/or the calcineurin/NFATc pathway occurs in aggressively advanced tumors (6,19,28,29).
We have reported that obese adipose tissue-related ER stress increases ANGPTL2 secretion or expression in adipocytes (13). ER stress is easily induced by stresses such as hypoxia, oxidative stress, hypoglycemia, and viral infection, all commonly observed in primary tumor microenvironment (30). We found Angptl2 mRNA levels in tumor cells were significantly increased under hypoxia and undernutrition ( Supplementary Fig. S15A). In addition, increased ANGPTL2 expression was detected in tumor cells in hypoxic regions ( Supplementary Fig. S15B), suggesting that tumor microenvironmental factors, such as hypoxia and/or undernutrition, induce ANGPTL2 expression in tumor cells. Cytoplasmic calcium concentrations increase due to ER stress-dependent calcium release from the ER (31) and activate the serine/threonine phosphatase calcineurin, which in turn dephosphorylates NFATc proteins and triggers their nuclear accumulation (32). NFATc function has been extensively studied in the immune system, but there is increased interest in NFATc activity in cancer (6).
We speculate that tumor cell-autonomous responses to the microenvironment, such as activation of the ER stress/calcineurin/NFATc pathway and/or Interestingly, a recent report suggests that ANGPTL2 might function in tumor refractoriness to anti-VEGF therapy (33). On the other hand, VEGF reportedly activates an inhibitor of calcineurin/NFATc signaling, namely, the Down syndrome critical region gene 1 (DSCR1) (34,35). Taken together with these two reports, our findings suggest that in tumors refractory to anti-VEGF treatment, suppression of VEGF signaling may inactivate DSCR1 and thereby activate NFATc, resulting in increased ANGPTL2 expression and ANGPTL2-dependent tumor progression. Overall, we propose that the tumor microenvironment activates NFATc and ATF/CREB family proteins in tumor cells, resulting in ANGPTL2 induction and subsequent tumor metastasis.
Based on the results of EMSA and ChIP assays, we suggested that NFATc, ATF2, and c-Jun form a complex and bind to the Angptl2 promoter region in PMA/Ion-treated H460 cells. CA-NFATc-dependent induction of the Angptl2 reporter activity was augmented by c-Jun co-expression, and the augmentation was significantly enhanced by ATF2 co-expression. In contrast, ATF2 alone did not induce the Angptl2 reporter activity. Therefore, we consider that ATF2 would be important to form a complex with NFATc and c-Jun to induce the Angptl2 reporter activity.
We previously reported that ANGPTL2 increases angiogenesis through Rac activation in endothelial cells (13). Consistently, here we found that tumor cell-derived ANGPTL2 increases tumor angiogenesis. Interestingly, we also found that ANGPTL2 directly accelerates tumor cell motility through Rac activation in tumor cells. We observed that EMT occurs prominently in   and low (n = 48) groups (P < 0.01 and P = 0.09 by log-rank test). *P < 0.05