Tumour stroma‐derived lipocalin‐2 promotes breast cancer metastasis

Tumour cell‐secreted factors skew infiltrating immune cells towards a tumour‐supporting phenotype, expressing pro‐tumourigenic mediators. However, the influence of lipocalin‐2 (Lcn2) on the metastatic cascade in the tumour micro‐environment is still not clearly defined. Here, we explored the role of stroma‐derived, especially macrophage‐released, Lcn2 in breast cancer progression. Knockdown studies and neutralizing antibody approaches showed that Lcn2 contributes to the early events of metastasis in vitro. The release of Lcn2 from macrophages induced an epithelial–mesenchymal transition programme in MCF‐7 breast cancer cells and enhanced local migration as well as invasion into the extracellular matrix, using a three‐dimensioanl (3D) spheroid model. Moreover, a global Lcn2 deficiency attenuated breast cancer metastasis in both the MMTV–PyMT breast cancer model and a xenograft model inoculating MCF‐7 cells pretreated with supernatants from wild‐type and Lcn2‐knockdown macrophages. To dissect the role of stroma‐derived Lcn2, we employed an orthotopic mammary tumour mouse model. Implantation of wild‐type PyMT tumour cells into Lcn2‐deficient mice left primary mammary tumour formation unaltered, but specifically reduced tumour cell dissemination into the lung. We conclude that stroma‐secreted Lcn2 promotes metastasis in vitro and in vivo, thereby contributing to tumour progression. Our study highlights the tumourigenic potential of stroma‐released Lcn2 and suggests Lcn2 as a putative therapeutic target. Copyright © 2016 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.


Introduction
Lipocalin-2 (Lcn2) is a 25 kDa glycoprotein of the lipocalin superfamily [1,2] that plays a pivotal role during bacterial infections [3], kidney regeneration [4], sepsis [5] and cancer [6]. Regarding tumour progression, several studies have indicated that Lcn2 expression correlates with poor prognosis [7][8][9]. Additionally, Lcn2 serves as a prognostic and diagnostic marker, because elevated levels of Lcn2 are detected in the urine of cancer patients [9]. Lcn2 displays pleiotropic functions and promotes proliferation, survival, differentiation and migration [10], thus rendering Lcn2 a putative mediator of tumour development. It was previously reported that Lcn2 promotes lung metastasis of murine breast cancer cells after injecting Lcn2-overexpressing 4 T1 cells [11]. Lcn2 has been proposed to promote early events of tumour metastasis. On the one hand, tumour-supporting effects of Lcn2 can be explained by stabilizing gelatinase B (MMP-9) [12], thereby enhancing degradation of the extracellular matrix and tumour cell dissemination [13,14]. On the other hand, Lcn2 induces epithelial-mesenchymal transition (EMT); specifically, overexpression of Lcn2 in MCF-7 breast cancer cells provokes EMT by reducing E-cadherin and increasing vimentin and fibronectin [9]. In contrast, the knockdown of Lcn2 in MDA-MB-231 breast cancer cells reverses EMT, associated with reduced tumour growth and metastasis [9]. In line with this, we recently described that Lcn2 conveys EMT characteristics to A375 melanoma cells, enhancing migration and invasion [15]. Moreover, the use of mouse mammary tumour Stroma-derived Lcn2 and metastasis 275 cells, whereas the possibility that Lcn2 is provided by tumour-infiltrating immune cells, such as neutrophils or macrophages, has not so far been taken into account.
Chronic inflammation and an impaired immune response provoke outgrowth of transformed cells and tumour progression [19]. Tumour-associated immune cells acquire a supportive phenotype to promote angiogenesis, metastasis and tumour cell proliferation. Tumour-associated macrophages (TAMs) are a prominent population of functionally polarized immune cells in the tumour micro-environment [20]. They infiltrate the majority of human tumours and are often linked to a poor prognosis [21]. TAMs not only contribute to primary tumour growth but also interact with tumour cells in the distinct phases of the metastatic route. There is strong evidence that migrating tumour cells co-localize with endothelial cells and macrophages in order to support metastatic spread [21][22][23]. The complex functional TAM phenotype is, at least in part, a response to tumour-derived components. We previously determined that apoptotic tumour cells activate the production and secretion of Lcn2 in macrophages, with the subsequent polarization of these macrophages towards a pro-tumour phenotype [4]. Along these lines, we recently showed that macrophage-derived Lcn2 promotes proliferation of MCF-7 breast cancer cells [24]. Furthermore, inhibition of Lcn2 production in macrophages reduced renal regeneration when applying a macrophage-based cell therapy approach in a renal ischaemia-reperfusion injury model, thereby substantiating the pro-proliferative and anti-inflammatory role of Lcn2 [4,25]. Taking  into  account  that  Lcn2  conveys pro-proliferative, pro-regenerative and anti-inflammatory properties, we hypothesized that breast cancer progression might rely, at least in part, on the presence of Lcn2 in the tumour-supportive microenvironment. We aimed at elucidating the role of macrophage-derived Lcn2 during the different stages of metastasis, including EMT, migration and invasion, both in vitro and in vivo.

Cell culture
The human breast cancer cell lines MCF-7 and MDA-MB-231, the hepatocellular carcinoma cell line HUH7 and the lung carcinoma cell line A549 were cultured in Dulbecco's modified Eagle's medium (DMEM) with high glucose (Life Technologies, Darmstadt, Germany), supplemented with 100 U/ml penicillin (PAA Laboratories, Cölbe, Germany), 100 μg/ml streptomycin (PAA Laboratories) and 10% fetal calf serum (FCS; PAA Laboratories). T47D human breast cancer cells were cultured in RPMI 1640 (Life Technologies), supplemented with 100 U/ml penicillin, 100 μg/ml streptomycin and 10% FCS. Tumour cells were cultivated in a humidified atmosphere with 5% CO 2 at 37 ∘ C and passaged three times weekly.

Generation of macrophage-conditioned medium
MDA-MB-231 cells were stimulated with 0.5 μg/ml staurosporine (LC Laboratories, Woburn, USA) for 1 h, washed with phosphate-buffered saline (PBS) and incubated overnight in RPMI to generate apoptotic-conditioned medium (ACM). Primary human macrophages were stimulated with ACM for 6 h to induce Lcn2 production, washed with PBS and cultured in RPMI overnight to generate macrophage-conditioned medium (MCM). To explore the role of Lcn2 in MCM, we applied a neutralizing antibody against Lcn2 (3.5 μg/ml; cat. no. MAB1757, R&D Systems, Wiesbaden, Germany). An isotype matching IgG antibody (3.5 μg/ml; cat. no. 6-001-A, R&D Systems) was used as a control.

Production of recombinant Lcn2
Recombinant human Lipocalin-2 (LCN2) was produced by transformation of Escherichia coli with a pGEX-4 T-3-NGAL plasmid (kind gift from Dr Anna Sola, IIBB, Barcelona, Spain) expressing human LCN2 tagged to glutathione-S-transferase. LCN2 expression was initiated by supplementing isopropyl-β-D-thiogalactopyranoside (Sigma-Aldrich, Steinheim, Germany) to the bacterial culture to activate the lac operon. LCN was purified using ProCatch Glutathione Resin (Miltenyi Biotec, Bergisch Gladbach, Germany).  and lung were isolated. Lung metastases were determined using Meyer's haemalum (Merck, Darmstadt, Germany) staining. The appearance of metastases was evaluated in 12 lung sections/mouse at different levels and the percentage of mice with lung nodules was calculated. Mice with an age of at least 20 weeks were taken into account.

Xenograft model
MCF-7 cells were pretreated with MCM (scRNA MCM) and MCM from Lcn2-deficient macrophages (siLcn2 MCM) for 3 days. Subsequently, 3 × 10 6 tumour cells were resuspended in 25 μl PBS/25 μl growth factor-reduced Matrigel (Corning, New York, USA) and injected into the mammary fat pad of BALB/c nude female mice; each mouse received two plugs. Tumour growth was monitored to the size of 0.3 cm 3 . Oestrogen was provided by subcutaneous implantation of oestrogen pellets (90-day release) (Innovative Research of America, USA) immediately after cell injections. Detection and quantification of disseminated tumour cells in the lung was performed using the IVIS-200 system [26]. Lung sections were stained with haematoxylin and eosin (H&E). For immunohistochemistry, a rabbit monoclonal antibody to human oestrogen receptor (ER)α (ab16660, Abcam) and a mouse monoclonal antibody to human Ki-67 (NCL-L-Ki-67-MM1, Novocastra) were used.

RNA extraction and reverse transcription-quantitative PCR (RT-qPCR)
RNA extraction, reverse transcription and qPCR were performed as previously described [24]. The human primers used were:

Stroma-derived Lcn2 and metastasis 277
Transendothelial migration assay MCF-7 cells were stimulated with 1 μg/ml Lcn2 or MCM for 4 days and marked with Cell Tracker green. HPMECs (1 × 10 5 ) were seeded into fibronectin-coated Transwell inserts of a 24-well plate (Greiner Bio-One, Frickenhausen, Germany) and grown to confluence. MCF-7 cells (5 × 10 4 ) were added into each insert. After 24 h, the cells were washed and fixed with 4% PFA. Five fields/well were counted and the mean was calculated for each group. At least three independent experiments using triplicates were performed.
Cell migration assay MCF-7 cells were transfected either with siSLC22A17 or scrambled control siRNA (scRNA) and a cell migration assay was performed, as described previously [24]. Measurement of the scratch area was accomplished using ImageJ software. The 0 h value was normalized to the 24 h value of each group. The relative migration rate defines the mean value of at least three independent experiments using triplicates.

Spheroid invasion assay
MCF-7 spheroids were generated in 1.5% agarosecoated 96-well-plates (Greiner Bio-One). Spheroids were stimulated with scRNA MCM and siLcn2 MCM every 3 days for 2 weeks and subsequently embedded in a collagen I matrix for 7 days. Three images were taken of each spheroid from each group (triplicates) from at least three independent experiments. Distance of invasion from the spheroid border into the collagen I matrix was measured using Axiovision software (Zeiss).

Statistical analysis
Each experiment was performed at least three times.
The p values were calculated using Student's t-test or one-way ANOVA and considered significant at *p < 0.05, **p < 0.01 or ***p < 0.001.

Lcn2 ablation delays tumour growth and metastasis of PyMT breast tumours
We analysed the impact of Lcn2 in the spontaneous PyMT breast cancer model, where all female mice developed a tumour within 150 days in at least one mammary gland. We detected a significant delay of tumour growth in Lcn2 −/− PyMT mice, starting at week ∼16, whereas tumour growth in wt PyMT mice was already detected at week ∼12 ( Figure 1A). Accordingly, Lcn2 −/− PyMT mice reached the end point of 1.5 cm tumour diameter at later times (week ∼23), compared to wt PyMT mice that were sacrificed at week ∼19 ( Figure 1B). In order to define the extent of overall tumour development, we chose 18 weeks, a time-point when all mice showed tumour burden. Lcn2 −/− PyMT mice had smaller tumours and showed more mammary glands without tumour burden compared to wt PyMT mice ( Figure 1C), but the tumour diameter and number of glands with tumours did not significantly differ at the end point of 1.5 cm ( Figure 1D). Interestingly, we saw significant differences in the number of metastasis-bearing mice, comparing wt and Lcn2 −/− groups at sacrifice. The percentage of mice developing lung metastases was significantly lower in the Lcn2 −/− group compared to the wt group ( Figure 1E, F).

Lcn2 induces metastatic spread in vitro
We previously described that not only tumour cells but also macrophages express Lcn2 after exposure to apoptotic tumour cell supernatants [4]. In the context of cancer, we reported that TAM-released Lcn2 supports tumour growth [24], but the impact on metastasis had not so far been investigated. Therefore, we generated macrophage-conditioned medium containing Lcn2 (MCM) by stimulation of primary human macrophages with apoptotic tumour cell supernatants (ACM) and used in vitro lung metastasis assays to characterize the ability of macrophage-released Lcn2 to promote the metastatic tumour cell phenotype. Most interestingly, ACM-stimulated macrophages secreted significantly higher levels of Lcn2 compared to tumour cells (see supplementary material, Figure S1A, B). The attachment of tumour cells to extracellular matrix (ECM) components, including collagen I and fibronectin, is an essential step for tumour cell dissemination into the lung. Both MCM-and recombinant Lcn2 (1 μg/ml)-stimulated cells significantly enhanced adhesion to both matrices. Quantification and representative pictures of adhesion to collagen I ( Figure 2A) and fibronectin ( Figure 2B) are displayed. Once tumour cells manage to adhere to the lung matrix, they migrate through the tight endothelial layer of the pulmonary tissue before a metastatic lesion can be established. To test the effect of Lcn2 on cancer cell migration across an endothelial layer, human pulmonary microvascular endothelial cells (HPMECs) were cultured in Transwell culture inserts. MCF-7 cells, prestimulated with MCM or recombinant human Lcn2 (1 μg/ml), passed more efficiently through this layer into the lower chamber of the Transwell ( Figure 2C). Thus, both MCM and Lcn2 treatment enhanced the passage of tumour cells through a pulmonary endothelial layer. As migration and invasion through the ECM are important events during metastasis, we examined the effect of macrophage-derived Lcn2 on tumour cell invasion, utilizing a previously described 3D tumour spheroid model [27]. To verify Lcn2 as the responsible factor in the MCM inducing invasiveness of breast cancer cells, we established a transient knockdown of Lcn2 in primary human macrophages in order to generate Lcn2-deficient MCM (siLcn2 MCM). Control macrophages were treated with scRNA in order to generate supernatants containing Lcn2 (scRNA MCM) (see supplementary material, Figure S2). MCF-7 spheroids were stimulated with siLcn2 MCM or scRNA MCM or left in normal growth medium (control) for 10 days. Subsequently, spheroids were embedded into a collagen I matrix for an additional 7 days to allow invasion into the matrix. ScRNA MCM-treated spheroids showed significantly enhanced invasion into the collagen I matrix, whereas siLcn2 MCM-stimulated spheroids displayed reduced invasion ( Figure 2D).

TAM-released Lcn2 facilitates lung colonization in vivo
As we had established that macrophage-derived Lcn2 induces metastatic spread in vitro, we wondered whether these effects could be transferred to a breast cancer in vivo model. Therefore, we employed a xenograft model using MCF-7 cells that were pretreated with scRNA MCM and siLcn2 MCM for 3 days. Our analysis of the tumour volume did not detect differences in tumour growth ( Figure 3A). Also, the weight of tumours at sacrifice was equivalent ( Figure 3B). Interestingly, the dissemination of implanted MCF-7 tumour cells to the lungs was significantly decreased by siLcn2 MCM-pretreatment ( Figure 3C). Furthermore, histological analysis was performed in order to detect ER-positive proliferating (Ki-67-positive) tumour cell colonies ( Figure 3D). Representative pictures are shown for the co-localization of ER-and Ki-67-positive treated with scRNA MCM containing Lcn2 or siLcn2 MCM deficient of Lcn2 or remained unstimulated (control) for 10 days and were embedded into a collagen I matrix for an additional 7 days. The distance of spheroid cell invasion from the spheroid border into the matrix was measured after 1, 3 and 7 days and is represented as distance of invasion relative to day 0; representative pictures are shown. Scale bar = 200 μm; data are mean ± SD; n ≥ 3; *p < 0.05, **p < 0.01 and ***p < 0.001 versus controls metastatic lesions. These data support the hypothesis that macrophage-derived Lcn2 increases the invasiveness of breast cancer cells, thereby facilitating metastatic spread.

Macrophage-derived Lcn2 induces EMT through its specific receptor SLC22A17
Since the higher migratory and invasive character of tumour cells is often associated with the transition from an epithelial to a mesenchymal state, the process of EMT plays an integral role in cancer metastasis. Therefore, we reasoned that macrophage-derived Lcn2 promotes EMT in breast cancer cells. MCM elicited a decrease in transcript abundance for the epithelial marker E-cadherin (CDH1) ( Figure 4A) and an increase of the mesenchymal marker N-cadherin (CDH2) ( Figure 4B) in human MCF-7 breast cancer cells. This effect was significantly attenuated by neutralizing Lcn2, while addition of the IgG control antibody was without effect. To exclude potential cell-specific effects, we verified changes in CDH1 and CDH2 mRNA abundance in response to MCM in T47D breast cancer cells, A549 lung carcinoma cells and HUH7 hepatocellular cells (see supplementary material, Figure S3). In all cell lines tested, MCM decreased CDH1 and increased CDH2 mRNA expression. These changes were significantly reversed by neutralizing Lcn2. Analysis of EMT-related transcription factors showed the potential involvement of ZEB1, whereas SNAI1 was not affected (see supplementary material, Figure S4). Neither Slug (SNAI2) nor Twist (TWIST1)  Figure S5). Both control (scRNA) and SLC22A17-down-regulated tumour cells were stimulated with MCM for 24 h and the abundance of mRNA encoding the EMT markers E-cadherin ( Figure 4C) and N-cadherin ( Figure 4D) were measured by RT-qPCR. The previously observed impact of MCM on EMT marker expression was significantly reversed by the knockdown of SLC22A17. Since EMT correlates with increased migration rates, we explored whether the knockdown of SLC22A17 affected tumour cell migration. Indeed, the knockdown of SLC22A17 did impair tumour cell migration ( Figure 4E). Quantification showed a significantly reduced migration rate in siSLC22A17 knockdown cells ( Figure 4F).

Stroma-derived Lcn2 induces breast cancer cell dissemination into the lung
In order to evaluate the impact of stroma-derived Lcn-2 on tumour growth and metastasis, we established an orthotopic breast cancer model. We used both wt and Lcn2 −/− PyMT mice as tumour cell donors and Lcn2 +/+ and Lcn2 −/− C57BL/6 mice as recipients. A schematic picture of the model is shown in Figure 5A. Implantation of Lcn2 −/− tumour cells into Lcn2 +/+ and Lcn2 −/− mice allowed us to analyse stroma-derived Lcn2 effects on tumour growth and metastasis. In addition, the comparison of implanted wt and Lcn2 −/− tumour cells allowed us to determine tumour cell-derived Lcn2 effects. We could not detect any differences in tumour development between Lcn2 +/+ and Lcn2 −/− recipients for either Lcn2-competent donors ( Figure 5B) or Lcn2-deficient donors ( Figure 5C). Furthermore, the tumour weight at sacrifice did not differ significantly, suggesting that stroma-derived Lcn2 is not important for primary tumour growth ( Figure 5D-E). In order to detect lung metastases derived from implanted PyMT-positive tumour cells, we performed immunofluorescence using a PyMT specific antibody in combination with staining of proliferating tumour cells using Ki-67. Representative images show an efficient co-localization of PyMT-and Ki-67-positive tumour cells ( Figure 5F). Quantification of double-positive tumour cells showed that the injection of wt tumour cells into Lcn2 −/− recipient mice elicited a significant decrease in the number of disseminated Ki-67-positive PyMT tumour cells within the lung ( Figure 5G). However, implantation of Lcn2-deficient tumour cells resulted in a slight, but not significant, reduction of the number of disseminated  Figure 6B). Our results suggest that stroma-derived Lcn2 enhances metastatic spread by promoting different steps of the metastatic cascade, such as adhesion, transendothelial migration, invasion and EMT ( Figure 6C). cell dissemination by inducing EMT, resulting not only in increased cancer cell motility and invasion but also in enhanced transendothelial migration. A number of studies previously acknowledged a role of tumour-derived Lcn2 in tumour development and metastatic breast cancer progression, correlated with a poor prognosis [9,16]. The present study adds to the emerging role of Lcn2 for tumour progression. Interestingly, our data suggest a significantly higher expression of Lcn2 in ACM-treated macrophages than in tumour cells (see supplementary material, Figure S1). It was previously suggested that a threshold level of Lcn2 must to be achieved in tumour cells in order to promote EMT in Lcn2-overexpressing MCF-7 cells [9]. Therefore, it might be speculated that the local expression of Lcn2 from tumour-infiltrating macrophages adds to their pro-tumorigenic capacity. So far, most studies have focused on the role of tumour cell-derived Lcn2, whereas the idea that Lcn2 might originate from tumour-infiltrating immune cells has so far not been appreciated. Shinriki et al [28] recently analysed the expression of Lcn2 in human oral squamous cell carcinoma (OSCC); they detected that poorly differentiated OSCC and reduced overall survival were particularly associated with Lcn2-expressing, CD68-positive stromal immune cells, rather than tumour cells, thus underscoring the importance of monocyte-derived Lcn2. These findings point to a role of stroma-derived Lcn2 as a previously unrecognized factor in tumour progression. However, we previously showed that macrophage-released Lcn2 enhanced the proliferative capacity of MCF-7 breast cancer cells in vitro [24]. To prove the impact of macrophage-derived Lcn2 on breast cancer progression in vivo, we used a xenograft model involving MCF-7 cells pretreated with macrophage supernatants and an orthotopic mammary tumour model. Neither the xenograft ( Figure 3A, B) nor the orthotopic model (5B-E) displayed differences in primary tumour growth. In the present study we also checked for tumour growth, both in a 3D tumour spheroid model and in 2D cell culture, but failed to detect differences in primary growth characteristics in the complex 3D system, whereas the 2D-cultured cells responded as described previously [24]. Corroborating our data, it was recently reported [29] that supernatants of macrophages co-cultured with apoptotic tumour cells promote the aggressiveness of MCF-7 cells by enhancing tumour growth and metastasis.
There is growing evidence that Lcn2 promotes the development of malignant tumour cell phenotypes by inducing EMT [9,15], an essential early step towards tumour metastasis. The transition from an epithelial into a mesenchymal cell is characterized by the loss of epithelial cell polarity, the loss of cell-cell contacts and increased cell motility. An important characteristic of EMT is the cadherin switch, characterized by decreased E-cadherin but increased N-cadherin expression. E-cadherin promotes cell-cell adhesion through homophilic interactions between E-cadherin proteins on adjacent cells, thus forming adherens junctions [30]. In contrast, N-cadherin causes tumour progression and invasion. This is facilitated through interactions with fibroblast growth factor receptor 1, activating the mitogen-activated protein kinase (MAPK)-extracellular signal-regulated kinase 284 B Ören et al (ERK) pathway to enhance MMP-9 expression [30,31]. Interestingly, Lcn2 was shown to support tumour progression by stabilizing MMP-9, thereby facilitating ECM degradation [12,14]. Apparently, tumour cell-derived Lcn2, as well as stroma-derived Lcn2, might be important players during metastatic spread. However, our orthotopic tumour model suggests that predominantly stroma-derived Lcn2 induces metastasis by promoting tumour cell dissemination into the lung. Therefore, we questioned whether this results from the induction of EMT in tumour cells. In fact, we showed that macrophage-derived Lcn2 induces EMT in MCF-7 breast cancer cells by reducing the epithelial marker E-cadherin and inducing the mesenchymal marker N-cadherin ( Figure 4A-D). Importantly, we observed a more robust and consistent Lcn2-induced expression of N-cadherin rather than repression of E-cadherin in vitro. However, it is important to note that EMT is usually not a complete transition, but rather a transient and reversible process, since many tumours might also co-express both cadherins, thereby illustrating their plasticity [32]. Nevertheless, it is speculated that a proportion of fully acquired EMT cells are required for effective invasion through the ECM, thereby opening the way for non-or transient-EMT cells to enter the bloodstream and to spread into distant organs.
In summary, we propose that stroma-derived Lcn2 enhances the malignant characteristics of breast cancer cells. Our studies underscore the significance of stroma-derived Lcn2 on tumour cell dissemination and metastatic growth and offer new therapeutic perspectives. Nevertheless, further research is needed to define the diverse biological effects of Lcn2 within the tumour micro-environment.

SUPPLEMENTARY MATERIAL ON THE INTERNET
The following supplementary material may be found in the online version of this article: Figure S1. Release of Lcn2 from macrophages and tumour cells