Lipocalin 2 promotes inflammatory breast cancer tumorigenesis and skin invasion

Inflammatory breast cancer (IBC) is an aggressive form of primary breast cancer characterized by rapid onset and high risk of metastasis and poor clinical outcomes. The biological basis for the aggressiveness of IBC is still not well understood and no IBC‐specific targeted therapies exist. In this study, we report that lipocalin 2 (LCN2), a small secreted glycoprotein belonging to the lipocalin superfamily, is expressed at significantly higher levels in IBC vs non‐IBC tumors, independently of molecular subtype. LCN2 levels were also significantly higher in IBC cell lines and in their culture media than in non‐IBC cell lines. High expression was associated with poor‐prognosis features and shorter overall survival in IBC patients. Depletion of LCN2 in IBC cell lines reduced colony formation, migration, and cancer stem cell populations in vitro and inhibited tumor growth, skin invasion, and brain metastasis in mouse models of IBC. Analysis of our proteomics data showed reduced expression of proteins involved in cell cycle and DNA repair in LCN2‐silenced IBC cells. Our findings support that LCN2 promotes IBC tumor aggressiveness and offer a new potential therapeutic target for IBC.

Inflammatory breast cancer (IBC) is an aggressive form of primary breast cancer characterized by rapid onset and high risk of metastasis and poor clinical outcomes. The biological basis for the aggressiveness of IBC is still not well understood and no IBC-specific targeted therapies exist. In this study, we report that lipocalin 2 (LCN2), a small secreted glycoprotein belonging to the lipocalin superfamily, is expressed at significantly higher levels in IBC vs non-IBC tumors, independently of molecular subtype. LCN2 levels were also significantly higher in IBC cell lines and in their culture media than in non-IBC cell lines. High expression was associated with poor-prognosis features and shorter overall survival in IBC patients. Depletion of LCN2 in IBC cell lines reduced colony formation, migration, and cancer stem cell populations in vitro and inhibited tumor growth, skin invasion, and brain metastasis in mouse models of IBC. Analysis of our proteomics data showed reduced expression of proteins involved in cell cycle and DNA repair in LCN2silenced IBC cells. Our findings support that LCN2 promotes IBC tumor aggressiveness and offer a new potential therapeutic target for IBC.

Introduction
Inflammatory breast cancer (IBC) is the most aggressive and deadly variant of primary breast cancer. Although IBC is considered rare in the United States (1-4% of all breast cancer cases), it accounts for a disproportionate 10% of breast cancer-related deaths because of its aggressive proliferation and metastasis and limited Abbreviations BrMS, brain metastasis subline; ER, estrogen receptor; ERBB2/HER2, Erb-B2 receptor tyrosine kinase 2; IBC, inflammatory breast cancer; LCN2, Lipocalin 2; LuMS, lung metastasis subline; MAPK, mitogen-activated protein kinase; MMP-9, matrix metallopeptidase 9; MTOR, mammalian target of rapamycin; non-IBC, non-inflammatory breast cancer; PR, progesterone receptor; RPS6KB1, ribosomal protein S6 kinase B1; shCtl, short hairpin RNA control; shLCN2, short hairpin RNA for Lipocalin 2; TNBC, triple-negative breast cancer. therapeutic options [1][2][3][4][5]. IBC disproportionately affects young and African American women [1,6]. IBC is associated with unique clinical and biological features and a distinctive pattern of recurrence with high incidence in central nervous system, lung, and liver as first site of relapse [4,7,8]. Even with multimodality treatment strategies, survival rates for women with IBC are far lower than for those with other types of breast carcinoma (non-IBC), with estimated 5-year overall survival rates limited to 40% vs 63% for non-IBC [4,[6][7][8][9]. These features underscore the critical need to better define the mechanisms that drive the aggressive behavior of IBC and to develop novel agents to improve the overall prognosis for women with IBC. Efforts have been undertaken to identify pathways and therapeutic targets distinct to IBC and to better elucidate the mechanisms of IBC aggressiveness [10][11][12][13][14][15]. However, the molecular and cellular basis for IBC aggressiveness remains unclear. Identification of specific targets and unraveling the mechanisms of growth and metastasis of this aggressive disease could lead to improvements in IBC patient survival.
In the present study, we demonstrate that LCN2 was expressed at significantly higher levels in patients with IBC and that LCN2 promoted tumor growth, skin invasion, and metastasis in xenograft mouse models of IBC.

ELISA
Human LCN2/NGAL Quantikine ELISA Kits (#DLCN20; R&D Systems, Minneapolis, MN, USA) were used to measure the levels of LCN2 in the cell lines according to the manufacturer's instructions. Samples were assayed in duplicate.

Proliferation
About 2500 cells were seeded in triplicate in a 96-well plate. Cell proliferation was measured every day for up to 72 h with the CellTiter-Blue assay (#G8080; Promega, Madison, WI, USA) according to the manufacturer's instructions. Absorbance was recorded at OD595 nm with a Multifunctional Reader VICTOR X 3 (PerkinElmer, Waltham, MA, USA).

Colony-formation assay
About 100 SUM149 or 500 MDA-IBC3 shRNA Control or LCN2-silenced cells were plated in triplicate in 6-well plates. After 15 days, cells were fixed with methanol for 2 min and stained with 0.2% (w/v) crystal violet for 30 min. Colonies were counted using Gel-Count (Oxford Optoronix, Abingdon, UK).

Migration and invasion assay
For the migration assay, 50 000 cells per well (triplicate) were seeded in medium without serum onto 8 lm polypropylene filter inserts in Boyden chambers (Fisher). Medium with 10% FBS was added onto the well. After 24 h, cells on the bottom of the filter were fixed and stained with Thermo Scientific Shandon Kwik Diff Stains (Fisher). The invasion assay was done as described above, except that the 8 lm polypropylene filter inserts were coated with Matrigel (#CB-40234; Corning Inc., Corning, NY, USA) and incubated for 24 h. Ten visual fields were randomly chosen under microscopy and cells were quantified by using IMAGEJ software (National Institutes of Health, Bethesda, MD, USA).

Kinase enrichment analysis
The RPPA data were also used for the phosphoproteomic analysis using kinase enrichment analysis (KEA -https://maayanlab.cloud/kea3/) [31]. Briefly, the 20 proteins that exhibit the highest phosphorylation fold change levels in control vs LCN2-silenced cells were analyzed. Two different analyses were performed using KEA: (a) The differentially phosphorylated proteins are queried for enrichment of kinase substrates and (b) the differentially phosphorylated proteins are queried for enrichment of interacting proteins across seven databases. The latter analysis is more general and is not limited to only kinase substrates. Both analyses result in the detection of kinases that are putatively responsible for the observed phosphorylation differences. Identified proteins by both analyses were mapped onto the STRING network (https://string-db. org) to investigate their mutual interactions.

In vivo experiments
Four-to six-week-old female athymic SCID/Beige mice were purchased from Harlan Laboratories (Indianapolis, IN, USA) and allowed to acclimate for 1 week before use. All mice were given free access to food and water in a specific pathogen-free condition. All animal experiments were done in accordance with protocols approved by the Institutional Animal Care and Use Committee of MD Anderson Cancer Center. Mice were euthanized with overdose of isoflurane when they met the institutional criteria for tumor size and overall health condition. For primary tumor growth, cells were injected into the orthotopic cleared mammary fat pad of mice as previously described [32]. Briefly, 5 9 10 5 SUM149 shRNA Control / LCN2-knockdown cells were injected (9 mice/Control; 10 mice/LCN2 KD). Tumor volumes were assessed weekly by measuring palpable tumors with calipers. Volume (V) was determined as V = (L 9 W 9 W) 9 0.5, with L being length and W width of the tumor. To determine latency, the first day when palpable tumors appeared was used to plot the graph. For brain metastatic colonization studies, we followed our laboratory protocol [33]. Briefly, 1 9 10 6 MDA-IBC3 GFP-labeled shRNA Control/LCN2knockdown cells (10 mice/group) were injected via the tail vein into SCID/Beige mice. At 12 weeks after tailvein injection, mice were euthanized, and brain tissue collected and imaged with fluorescent stereomicroscopy (SMZ1500; Nikon Instruments, Melville, NY, USA). IM-AGEJ was used to measure GFP-positive areas to quantify the area of brain tumor burden. For mice with more than one brain metastasis, the area of each metastasis was considered and measured.

Statistical analysis
All in vitro experiments were repeated at least three times, and graphs depict mean AE SEM. Statistical significance was determined with Student's t-tests (unpaired, two-tailed) unless otherwise specified. One-way analysis of variance was used for multiple comparisons. Mann-Whitney test was used when normality was not met. LCN2 expression in breast cancer samples was analyzed in the IBC Consortium dataset [34] for IBC and from a meta-dataset previously published [35]. Tumor samples were stratified as LCN2-high when expression in tumor was at least two-fold the mean expression level measured in the normal breast samples; otherwise, the sample was classified as LCN2low. Kaplan-Meier curves and log-rank tests were used to compare survival distributions. Univariate and multivariate Cox regression models were used to evaluate the significance of LCN2 expression on overall survival. A P value of < 0.05 was considered significant. GRAPHPAD software (GraphPad Prism 8, La Jolla, CA, USA) was used.

LCN2 mRNA is highly expressed in inflammatory breast cancer
Previous studies have shown that high LCN2 expression levels were correlated with poor prognosis in breast cancer patients [17,[25][26][27]. We further validated these findings by analyzing a meta-dataset of 8951 breast cancers, in which 87% of tumor samples were classified as LCN2-low (n = 7830/8951) and 13% as LCN2-high (n = 1121/8951). Table 1 summarizes the clinico-pathological patient characteristics stratified by LCN2 expression status. High expression of LCN2 was associated with variables commonly associated with poor outcome: younger patients' age, high grade, advanced stage tumors (pN-positive and pT3), ductal type, estrogen receptor (ER)-negative status, progesterone receptor (PR)-negative status, Erb-B2 receptor tyrosine kinase 2 (ERBB2)-positive status, and aggressive molecular subtypes [ERBB2 + and triple-negative breast cancer (TNBC) subtypes]. In this cohort, we also analyzed the association of LCN2 expression and survival over time using the Kaplan-Meier method. We found that LCN2-high tumors had significantly shorter overall survival (P < 0.0001) than LCN2-low tumors (Fig. 1A).
Analysis of microarray data from the IBC World Consortium Dataset [34] consisting of IBC and non-IBC patient samples (n = 389; IBC = 137, non-IBC = 252) showed that LCN2 expression was significantly higher in tumors from IBC patients compared to non-IBC (P = 0.0003; Fig. 1B). We validated this finding in another independent dataset [36] that compared mRNA expression of microdissected IBC and non-IBC tumors (P = 0.0379; Fig. 1C). Here too, LCN2 expression was higher in ER-negative IBC patients compared to ER-positive (P = 0.0009; Fig. 1D) and in more aggressive subtypes, ERBB2positive and TNBC, compared to hormone receptor (HR)-positive/ERBB2-negative subtype (Fig. 1E). Multivariate analysis showed that LCN2 was expressed significantly higher in IBC tumors relative to non-IBC tumors, independently from the molecular subtype differences (Odds ratio, 1.71, P = 0.034; Table 2). Here too, the survival analysis in IBC patients showed that LCN2-high tumors had significantly shorter overall survival (P = 0.0317) than LCN2-low tumors (Fig. 1F). Consistent with the patient data, the levels of LCN2 were higher in IBC cell lines (Fig. 1G) and in the supernatants collected from IBC cell lines relative to non-IBC (Fig. 1H). Taken together, our findings show that LCN2 is highly expressed in IBC tumors and is correlated with aggressive features and poor outcome suggesting it may contribute to the aggressive pathobiology of IBC tumors.

LCN2-knockdown reduced aggressiveness features in vitro
We generated stable LCN2-knockdown cell lines [SUM149 (triple-negative IBC); MDA-IBC3 (HER2 + IBC)] to investigate the role of LCN2 in IBC aggressiveness in vitro and in vivo. LCN2-knockdown was confirmed by qRT-PCR and immunoblotting ( Fig. 2A,  B). Because LCN2 is a secreted protein, we evaluated levels of LCN2 protein in the supernatants from control and LCN2-silenced IBC cell lines by using ELISA. We observed significant reduction of secreted LCN2 in the LCN2-silenced IBC cells (Fig. 2C). Silencing LCN2 slightly reduced proliferation of SUM149 cells but did not affect MDA-IBC3 cells (Fig. 2D). Depletion of LCN2 reduced the capacity of the cells to form colonies (Fig. 2E) and to migrate and invade (Fig. 3A,  B). LCN2 silencing also significantly reduced the percentage of cancer stem cell populations in LCN2silenced IBC cells relative to control, as shown by reductions in primary and secondary mammosphere formation efficiency (Fig. 3C,D) and CD44 + CD24 À cell subpopulations (Fig. 3E). These findings indicate  that suppression of LCN2 in IBC cells reduced in vitro aggressiveness features.

Silencing of LCN2 inhibited tumor growth and skin invasion
To investigate the effects of LCN2 on tumor growth and skin invasion, key characteristics of IBC tumors [4], we injected SUM149 control or LCN2-silenced cells into the cleared mammary fat pad of SCID/Beige mice. Silencing of LCN2 reduced tumor volumes (P = 0.0037; Fig. 4A) and tumor latency, that is, the ability to initiate tumor growth: mice transplanted with SUM149 LCN2-silenced cells took longer to initiate tumors than did those transplanted with SUM149 control cells (P = 0.0145; Fig. 4B). Because IBC typically manifests with skin invasion and formation of tumor emboli [4], we assessed skin invasion visually during primary tumor growth, as evidenced by loss of fur at the tumor site and skin redness and thickness, and during tumor excision when tumors were firmly connected with the skin. Analysis of resected tumors showed that significantly fewer mice with SUM149 LCN2-silenced cells had skin invasion/recurrence compared with mice implanted with control cells [shLCN2: two of eight mice (25%) vs shControl: seven of eight mice (87.5%), P = 0.01; Fig. 4C,D]. On histologic examination, tumors generated from LCN2-silenced cells were more differentiated than those generated from control SUM149 cells (Fig. 4E); we further  observed tumor emboli, another hallmark of IBC tumors, in SUM149 control-transplanted tumors but not in tumors generated from LCN2-silenced SUM149 cells (Fig. 4E). We recently generated xenograft mouse models of brain and lung metastasis via tail-vein injection of IBC cell lines [29,33]. We also showed that sublines of SUM149 generated from brain metastases (BrMS) and lung metastases (LuMS) have distinct morphologic and molecular features [29]. Microarray profiling of these sublines showed upregulation of LCN2 in the brain metastatic sublines (Fig. S1A), and we confirmed higher levels of secreted LCN2 in the BrMS sublines vs LuMS by ELISA (Fig. S1B). Most recently, Chi et al. [37] elegantly demonstrated that LCN2 promotes brain metastatic growth in mouse models of leptomeningeal metastasis, highlighting a potential brain metastasis-promoting role for LCN2. We investigated the functional role of LCN2 in IBC brain metastasis by using our HER2 + MDA-IBC3 mouse model, which has a high propensity to metastasize to the brain and has been used to identify targets and develop therapeutics against brain metastasis [29,[38][39][40]. We found that the brain metastatic burden was significantly lower in mice that had received tail-vein injection of LCN2silenced MDA-IBC3 cells than in mice injected with control cells (Fig. 4F, P = 0.0059). Also, fewer mice injected with LCN2-silenced cells developed brain metastasis [one of 10 (10%)] than did mice injected with control cells [five of 10 mice (50%)], although this trend was not statistically significant (P = 0.1409; Fig. 4G). Representative stereofluorescence and hematoxylin and eosin images of brain metastasis are shown in Fig. 4H. Overall, our findings suggest that LCN2 may drive IBC tumor progression, skin invasion/recurrence, and brain metastasis.

LCN2 silencing impairs cell cycle-associated proteins
To identify potential mechanisms and pathways involved in suppression of tumor growth and skin invasion in LCN2-silenced cells, we used reverse-phase proteomics assay (RPPA) profiling to compare control and LCN2-silenced SUM149 cells. Our analysis showed reduced expression of cell cycle-associated proteins [such as AXL, FOXM1, Chk1, CDK1, Wee1, Aurora-B, and cyclin-B1 and the mammalian target of rapamycin (mTOR)/AKT pathway] in LCN2-silenced IBC cells (Fig. 5A). Gene set enrichment analysis revealed several key signaling pathways that were enriched in the control cells, including those associated with cell cycling, DNA repair, and mTOR signaling (Fig. 5B). Furthermore, we performed KEA [31] on the 20 proteins that exhibited the highest phosphorylation fold changes in LCN2control vs LCN2-silenced SUM149 cells (Table S1). Based on the set of predicted activated kinases (Tables  S1 and S2), an interaction network was generated (Fig. 5C). Based on the node degree distribution (i.e., the distribution of the number of interactions per gene in the network), mitogen-activated protein kinase 1 (MAPK1; N = 10), MAPK8 (N = 7), ribosomal protein S6 kinase B1 (RPS6KB1; N = 7), and MTOR (N = 11) appear to be central to LCN2 action in SUM149 cells.
Our immunoblotting experiments confirmed that silencing LCN2 reduces the phosphorylated forms of MEK (pMEK) and ERK (pERK) of the MAPK signaling pathway (Fig. 5D). Thus, LCN2 may regulate different pathways, including cell cycle, MAPK, and mTOR proteins to promote tumor growth in IBC.

Discussion
Inflammatory breast cancer is an aggressive form of breast cancer with poor survival outcomes. Although considerable effort has been undertaken to understand the unique biology of IBC, insights are still limited as to the molecular properties that mediate the development and aggressiveness of IBC. Herein, we report that the secreted glycoprotein LCN2 was highly expressed in tumors from IBC patients and in IBC cell lines. We further demonstrate, with in vitro and in vivo studies, that LCN2 has a tumor promoter function in IBC. LCN2 has been implicated in the progression of several types of human tumors. LCN2 expression is higher in solid tumors than in corresponding normal tissues [41,25], and it is mainly described as tumor promoter in many cancers, including pancreas, glioblastoma, thyroid, kidney, esophagus, and breast cancer [19,28,[42][43][44][45][46][47][48].
In breast cancer, increased LCN2 expression was associated with poor outcomes and shown to be an independent prognostic marker of disease-specific-free survival [27,49,48]. LCN2 also correlates with several important unfavorable prognostic factors in breast cancer, such as hormone-negative status, high proliferation levels, high histologic grade, and the presence of lymph node metastases [27,49,48]. Further, serum levels of LCN2 have been shown to correlate with cancer progression and higher likelihood of metastasis in  breast cancer [26,50]. The oncogenic role of LCN2 has been reported in xenograft and LCN2-knockout mouse models. Disruption of the LCN2 gene in MMTV-PyMT mice was found to suppress primary tumor formation without affecting lung metastasis [51]. Using the spontaneous MMTV-ErbB2(V664E) LCN2 À/À mouse model, Leng et al. [18] reported delayed tumor growth and reduced lung metastasis burden in these LCN2 À/À mice. Another group showed that injection of wild-type PyMT tumor cells into LCN2-deficient mice did not alter primary tumor formation but did significantly reduce lung metastasis [52]. LCN2 has also been shown to promote tumor progression in xenograft mouse models [16,26]. Consistent with these studies, our current work with xenograft mouse models of IBC supports that LCN2 has a tumor promoter function in IBC tumors. We demonstrated that silencing of LCN2 reduced tumor initiation and growth, skin invasion/recurrence, and brain metastasis burden in preclinical mouse models of IBC. We further reported that depletion of LCN2 in IBC cell cultures reduced features associated with aggressiveness in vitro, including migration, invasion, and cancer stem cell populations. Others have also found that reduction of LCN2 levels affected the same features in MDA-MB-231 cells (TNBC cell line) and in SK-BR-3 (HER2 + breast cancer cell line) [18,26]. However, our data demonstrating higher levels of secreted LCN2 in IBC vs non-IBC cell lines and showing significant inhibition of key IBC tumor features such as tumor emboli/skin invasion in LCN2-silenced tumors suggest that LCN2 may exert its influence via an IBCspecific mechanism. The LCN2 protein has many functions, including transport of fatty acids and iron, induction of apoptosis, suppression of bacterial growth, and modulation of inflammatory responses [16][17][18][19][20]26,53]. In malignant cells, LCN2 promotes oncogenesis through several mechanisms, including stabilization of MMP-9, sequestration of iron, induction of EMT, apoptosis resistance, and regulation of cell cycling [16][17][18][19][20]26,53]. Here, we report that LCN2 could regulate cell cycle-associated proteins such as FOXM1, Chk1, CDK1, Aurora-B, Wee1, and cyclin-B1 to promote its oncogenic role in IBC tumors. Others have also found that silencing of LCN2 affected the expression of cell cycle proteins by reducing cyclin-D1 and inducing p21, resulting in G0-G1 cell cycle arrest [22,24,25].
LCN2 is also a potential therapeutic target in cancer and other diseases. An antibody against LCN2 was found to decreased lung metastasis in a 4T1-induced aggressive mammary tumor model [18]. In cervical cancer cells, treatment with LCN2-neutralizing antibody reduced the migration and invasion of cells that overexpressed LCN2 [54]. In other diseases, use of an anti-LCN2-neutralizing antibody showed reductions in reperfusion injury after stroke and attenuated skin lesions in a psoriasis mouse model [55,56]. These findings suggest that LCN2 could be an exploitable therapeutic target in IBC and other aggressive tumors. Further studies are needed to explore therapeutic strategies in IBC models by using antibodies against LCN2 or targeting LCN2-associated molecular pathways, including those involved in cell cycling.

Conclusion
In summary, our studies provide evidence, for the first time, that LCN2 is highly upregulated in IBC tumors and that it is required for tumor growth and skin invasion in mouse models of IBC; our findings further suggest that LCN2 could be a therapeutic target for IBC and other aggressive cancers. part by the following grants: UPR/MDACC Partnership for Excellence in Cancer Research (U54CA096297-CA096300 to BGD), Susan G. Komen Career Catalyst Research Grant (CCR16377813 to BGD), American Cancer Society Research Scholar grant (RSG-19-126-01 to BGD), and State of Texas Morgan Welch Welch Inflammatory Breast Cancer Program.

Data accessibility
The data that support the findings of this study are available from the corresponding author (bgde-beb@mdanderson.org) upon reasonable request. Additional data are available as supplementary material.

Author contributions
ESV and BGD conceived and designed the project, performed most of the experiments, analyzed the data, and interpreted the results. XH, RL, WB, SRS, and KG performed some experiments. PF, FB, and XS helped with data analysis. JS provided statistical analysis support. SK provided pathological expertise and analysis of xenograft tumors. SVL, FB, GS-S, PV-M, NTU, WAW, and DT provided resources and contributed to revision of the manuscript. ESV and BGD wrote and edited the manuscript with input from all other authors.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. LCN2 expression is higher in sublines generated from brain metastasis (BrMS) than those generated from lung metastasis (LuMS). (A) Microarray analysis of sublines generated from BrMS or LuMS of SUM149 cells showed LCN2 to be one of the top upregulated genes in BrMS (red arrow). Samples are described in Debeb 2016 [29]. (B) LCN2 is secreted in higher levels in BrMS vs LuMS. Table S1. Top kinases predicted to be activated based on kinase-substrate interactions of differentially phosphorylated proteins. Table S2. Top kinases predicted to be activated based on kinase-substrate and protein-protein interaction analysis of differentially phosphorylated proteins across 10 different knowledge bases.