Estrogen Receptor (ER)α-regulated Lipocalin 2 Expression in Adipose Tissue Links Obesity with Breast Cancer Progression*

Background: Mechanisms underlying obesity-associated breast cancer incidence are incompletely understood. Results: Adipose tissue estrogen receptor (ER)α expression is inversely associated with adiposity and the expression and release of Lcn2, an adipokine promoting breast cancer cell proliferation and migration. Conclusion: ERα is critical for restraining adiposity and Lcn2 production. Significance: Modulation of adipose tissue ERα action is a potential approach to prevent obesity and reduce breast cancer risk. Obesity is associated with increased breast cancer (BrCA) incidence. Considering that inactivation of estrogen receptor (ER)α promotes obesity and metabolic dysfunction in women and female mice, understanding the mechanisms and tissue-specific sites of ERα action to combat metabolic-related disease, including BrCA, is of clinical importance. To study the role of ERα in adipose tissue we generated fat-specific ERα knock-out (FERKO) mice. Herein we show that ERα deletion increased adipocyte size, fat pad weight, and tissue expression and circulating levels of the secreted glycoprotein, lipocalin 2 (Lcn2), an adipokine previously associated with BrCA development. Chromatin immunoprecipitation and luciferase reporter studies showed that ERα binds the Lcn2 promoter to repress its expression. Because adipocytes constitute an important cell type of the breast microenvironment, we examined the impact of adipocyte ERα deletion on cancer cell behavior. Conditioned medium from ERα-null adipocytes and medium containing pure Lcn2 increased proliferation and migration of a subset of BrCA cells in culture. The proliferative and promigratory effects of ERα-deficient adipocyte-conditioned medium on BrCA cells was reversed by Lcn2 deletion. BrCA cell responsiveness to exogenous Lcn2 was heightened in cell types where endogenous Lcn2 expression was minimal, but components of the Lcn2 signaling pathway were enriched, i.e. SLC22A17 and 3-hydroxybutyrate dehydrogenase (BDH2). In breast tumor biopsies from women diagnosed with BrCA we found that BDH2 expression was positively associated with adiposity and circulating Lcn2 levels. Collectively these data suggest that reduction of ERα expression in adipose tissue promotes adiposity and is linked with the progression and severity of BrCA via increased adipocyte-specific Lcn2 production and enhanced tumor cell Lcn2 sensitivity.

Obesity is associated with increased breast cancer (BrCA) incidence. Considering that inactivation of estrogen receptor (ER)␣ promotes obesity and metabolic dysfunction in women and female mice, understanding the mechanisms and tissuespecific sites of ER␣ action to combat metabolic-related disease, including BrCA, is of clinical importance. To study the role of ER␣ in adipose tissue we generated fat-specific ER␣ knock-out (FERKO) mice. Herein we show that ER␣ deletion increased adipocyte size, fat pad weight, and tissue expression and circulating levels of the secreted glycoprotein, lipocalin 2 (Lcn2), an adipokine previously associated with BrCA development. Chromatin immunoprecipitation and luciferase reporter studies showed that ER␣ binds the Lcn2 promoter to repress its expres-sion. Because adipocytes constitute an important cell type of the breast microenvironment, we examined the impact of adipocyte ER␣ deletion on cancer cell behavior. Conditioned medium from ER␣-null adipocytes and medium containing pure Lcn2 increased proliferation and migration of a subset of BrCA cells in culture. The proliferative and promigratory effects of ER␣deficient adipocyte-conditioned medium on BrCA cells was reversed by Lcn2 deletion. BrCA cell responsiveness to exogenous Lcn2 was heightened in cell types where endogenous Lcn2 expression was minimal, but components of the Lcn2 signaling pathway were enriched, i.e. SLC22A17 and 3-hydroxybutyrate dehydrogenase (BDH2). In breast tumor biopsies from women diagnosed with BrCA we found that BDH2 expression was positively associated with adiposity and circulating Lcn2 levels. Collectively these data suggest that reduction of ER␣ expression in adipose tissue promotes adiposity and is linked with the progression and severity of BrCA via increased adipocyte-specific Lcn2 production and enhanced tumor cell Lcn2 sensitivity.
The alarming rise in obesity over recent decades is strongly associated with a concomitant increase in chronic disease incidence (1). It is generally appreciated that obesity promotes lowgrade systemic inflammation via metabolic dysfunction and immune cell activation (2)(3), and these factors, common underpinnings of type 2 diabetes and atherosclerosis, are also associated with increased prevalence of breast cancer (BrCA) 7 in women (4 -7).
Breast cancer is a leading cause of mortality in women, and the inability to predict, prevent, and treat metastatic breast cancer currently limits patient care. Obesity is linked with more aggressive forms of cancer with less favorable outcomes as obese women in the highest quintile of body mass index (BMI) have double the death rate from BrCA compared with lean counterparts (4). Although the mechanisms by which obesity and metabolic dysfunction increase BrCA risk remain unclear, recent studies have correlated altered levels of circulating factors including metabolites, hormones, adipokines, and cytokines/chemokines with increased BrCA cell proliferation and migration. In addition to visceral adipose previously linked with higher BrCA incidence, adipocytes are also the most abundant cell type of the mammary tumor stroma, and thus paracrine action on malignant epithelium is thought to impact early stages of carcinogenesis as well as responsiveness of established tumors to adjuvant therapies (8 -11).
Although many factors contribute to the development of metabolic dysfunction and obesity in human subjects, rare inactivating mutations in the estrogen receptor ␣ gene, ESR1, and common polymorphisms at this locus, are associated with adiposity, type 2 diabetes, atherosclerosis, and BrCA risk, independent of circulating hormone status (12)(13)(14)(15). Moreover, Dahlman-Wright and colleagues (16) showed that ESR1 expression is reduced in adipose tissue from obese women. Consistent with observations in human subjects, mice harboring a homozygous Esr1-null mutation manifest marked obesity, insulin resistance, and heightened tissue inflammation (17,18). Although a role for ER␣ in regulating metabolic homeostasis, adiposity, and insulin sensitivity is well established, the molecular targets of ER␣ action within glucoregulatory cell types remain incompletely understood.
Considering that ER␣ is markedly reduced in adipose from obese individuals and obesity elevates metastatic breast cancer risk in women, we investigated whether targeted loss of adipose tissue in ER␣ could promote increased adiposity, metabolic dysfunction, and a secretory profile promoting BrCA tumorigenesis. Herein we show that adipose tissue deletion of ER␣ increased adiposity and inflammation in female mice, driven in part by a marked elevation in the adipocyte-derived factor, lipocalin 2 (Lcn2). Conditioned media (CM) containing secreted factors from ER␣-deficient adipocytes or pure Lcn2 markedly increased proliferation and motility of a specific set of BrCA cell lines in culture. We found that BrCA responsiveness to exogenous Lcn2 was marked by differential expression of intrinsic Lcn2 signaling components. We identified that expression of 3-hydroxybutyrate dehydrogenase (BDH2), an enzyme responsible for the production of the mammalian siderophore 2,5-dihydroxybenzoic acid and critical for mitochondrial heme synthesis, as a critical component in determining cellular responsiveness to Lcn2. Importantly, we showed that expression of BDH2 in breast tumor biopsies was positively associated with obesity and circulating Lcn2 levels in women with BrCA. Our findings suggest that adipose tissue ER␣ expression is an important unifying link between obesity and breast cancer risk in women.

EXPERIMENTAL PROCEDURES
Animals-Male and female flox/flox (f/f) and adipose-specific ER␣ KO (FERKO) mice on a C57Bl6 background were generated by crossing ER␣ floxed mice (19) with transgenic lines in which Cre recombinase was driven by the aP2 (FABP4) promoter (20). Lep Ob mice were from Jackson Laboratories and maintained as previously described (18). The EAAE-ER␣ DNAbinding domain mutant mice were generated by the Korach laboratory (21,22) as previously described, and adipose tissue was harvested for subsequent qPCR analyses. Control or 17␤estradiol pellets (0.05 mg; 21 days, Innovative Research) were surgically inserted under the skin of Lep Ob mice and tissues were harvested after 21 days following a 6-h fast. Female mice from the UCLA hybrid mouse diversity panel (HMDP; supplemental Table S1), including 102 strains of inbred animals (23), were maintained on a high fat (HF)/high sucrose (HS) Western diet (Research Diets, D12266B) with the following composition, 16.8% kcal protein, 51.4% kcal carbohydrate, 31.8% kcal fat. Following fasting, animals were anesthetized with 4% isoflurane and exsanguinated prior to tissue harvest. Blood was collected into tubes containing EDTA, and plasma was separated by centrifugation. All procedures were performed in accordance with the Guide for Care and Use of Laboratory Animals of the National Institutes of Health, and approved by the Animal Research Committee of the University of California, Los Angeles.
Human Subjects-Pre-treatment tumor gene expression data were mined from breast cancer patients participating in the UCLA Translational Oncology Research International (TORI-B02) trial (24).
Body Composition-Female mice from the HMDP were measured for total body fat mass and lean mass by magnetic resonance imaging (MRI) using Bruker Minispec with software from Eco Medical Systems.
RNA Isolation and Expression Profiling in Adipose from HMDP Mice and BrCA Cell Lines-Total RNA was isolated from tissues using TRIzol (Invitrogen) according to the manufacturer's instructions. Total RNA was isolated from cell cultures using the Qiagen RNeasy columns according to the manufacturer's instructions. For microarrays, adipose tissue and BrCA cell (supplemental Table S2) RNA was hybridized to Affymetrix HT_MG-430A arrays and scanned using standard Affymetrix protocols. To reduce the risk of spurious association results, RNA normalization was performed after removing all individual probes with SNPs and all probesets containing 8 or more SNP-containing probes, which resulted in 22,416 remaining probesets.
Cell Isolations and Culture-HEK293 cells (ATCC) were maintained and passaged in DMEM containing 10% FBS. ER␣ stably expressing 3T3-L1 cells were generated by infecting 3T3-L1 pre-adipocytes with fresh retrovirus particles generated in PhoenixE cells transfected with pBABE containing the Esr1-ORF, or empty pBABE for control cells (pB). Stable transformants were selected for 1 week in puromycin (5 g/ml). For differentiation of 3T3-L1, cells were grown to confluence (day 0) before incubation in the standard DMI (dexamethasone, 3-methy-1-isobutylxanthine, insulin) differentiation mixture with the addition of rosiglitazone maleate (500 nM, Alexis Biochemicals) where indicated, for 3 days (day 3), then encouraged to lipid load for 4 days in the presence of 10 nM insulin (day 7). Primary white adipocytes were generated from stromal vascular fractions isolated from epididymal adipose tissue beds from both f/f Control and FERKO female mice as previously described (25,26). After isolation, cells were allowed to proliferate to confluence, then immediately differentiated as stated above for 3T3-L1 cells. Breast cancer cell lines were maintained as previously described (27). MCF7, T47D, ZR75-1, and EFM19 were all grown and maintained in RPMI with 10% FBS. Experiments with CM or Lcn2 were carried out in growth medium for 72 h or Hanks' balanced salt solution with 1% albumin for 24 h, respectively.
Conditioned Media and Lcn2 Incubation-HEK293 cells were cultured as above until ϳ70% confluent before transfection with the indicated amounts of HA-Lcn2 or GFP plasmid with Lipofectamine 2000, as per the manufacturer's instructions (Invitrogen). Cells were allowed to recover in growth medium for 24 h after which the medium was removed and replaced with serum-free medium containing 0.1% fatty acidfree BSA for 6 -8 h to generate the CM. Pure Lcn2 (R&D Systems) was reconstituted and bound to iron prior to addition to culture medium as described previously (28).
Breast Cancer Cell Proliferation and Migration-Proliferation and migration experiments were performed in growth medium, unless otherwise specified. For proliferation studies, cells were plated into 24-well culture plates at a density of 20,000 cells per well and allowed to attach overnight before the medium was replaced with the addition of either Lcn2 or GFP CM from HEK293A cells or media from Esr1-KD/D and Esr1-KD/Lcn2-KD cells at a ratio of 1:5 (CM:media). After 48 h cells were washed twice with PBS and stained for 15 min in Hoechst nuclear dye (Invitrogen). Cell number was estimated by plotting Hoechst fluorescence measured in each well, against a predetermined standard curve. Results were confirmed in separate experiments using a Coulter Particle Counter. For migration studies, cells were plated in 6-well plates at subconfluence and allowed to proliferate in the presence of the specified CM until ϳ95% confluent. Two standardized scratches were made through each well and images were acquired in at least three wells for each condition as previously described (29,30). Additional images were acquired 18 and 36 h later.
Quantitative RT-PCR-cDNA synthesis was performed on 1 g of total RNA using iScript cDNA synthesis kit (Bio-Rad) and qPCR were performed using iQ SYBR Green Supermix (Bio-Rad). Primer sequences for the specific target genes analyzed can be found in supplemental Table S3.
Plasmids and Constructs-Esr1-knockdown (KD) by short hairpin RNAs (shRNAs) cloned into the lentivirus vector pLKO.1-puro were purchased as lentiviral transduction particles from Sigma (MISSION; Esr1-B). Scrambled (Scr) shRNAs (non-target shRNA vector, catalog number SHC002V; Sigma), used for control experiments, contain a hairpin insert that generates siRNAs but contains 5 base pair mismatches to all known human and mouse genes. Particles were subsequently used to transduce 3T3-L1 preadipocytes (multiplicity of infection 10). Stably expressing cells were selected with puromycin (3 g/ml) for 1 week. Knockdown efficiency was assessed by qPCR and Western blotting.
ER␣ Expression Plasmid-A mouse Esr1-ORF cloning vector (OpenBiosystems) was used to subclone esr1 into pcDNA3.1 and delivered to cells using Lipofectamine 2000 as per the manufacturer's instructions (Invitrogen).
Lcn2 HA-tagged Expression Plasmid-Lcn2 cDNA was PCR amplified with the addition of attB/P Gateway sites from the pMSCV-Lcn2 plasmid (20) (a kind gift from Evan Rosen). The subsequent PCR fragment was gel purified and cloned into a customized N-terminal hemagglutinin (HA) pcDNA backbone vector (kind gift from Thomas Vallim, UCLA) using Gateway technology (Invitrogen), before use in transfection assays as described above. C/EBP plasmids were a kind gift from Stephen Farmer (Boston University). Poly(ADP-ribose) PPAR␥2 (peroxisome proliferator-activated receptor ␥ 2) plasmid was obtained from Peter Tontonoz (UCLA). All primers used for cloning are presented in supplemental Table S4.
Promoter Luciferase Constructs and Assays-The Lcn2 5Јupstream promoter (2.7 kb) was cloned by PCR from BACclone RP24-290P20 (CHORI) with the primers outlined in supplemental Table S4, designed to add extensions for Gateway BP-cloning and SacI and KpnI sites. PCR fragments were amplified by cloning into pDONR221 using Gateway technology (Invitrogen), transformed, expanded, and isolated by miniprep. Promoter containing plasmids and the luciferase expression plasmid (pGL4-basic, Promega) were separately double digested with SacI and KpnI for 1 h, then precipitated and washed by phenol/chloroform extraction. Digested DNA was resuspended in water and used in ligation reactions with T4 ligase at a ratio of 6:1 (insert: plasmid) according to the manufacturer's instructions (New England Biolabs) and transformed into competent bacteria. Positive clones were sequenced for confirmation. For luciferase assays, 3T3-L1 cells were transfected with promoter plasmids or empty plasmid (500 ng) and Renilla (1 ng) together with other expression plasmids as indicated (100 ng each) using PLUS-reagent and Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Cells were then placed in phenol red-free DMEM containing 10% charcoal/dextran-stripped FBS for 12 h to recover before treatment with either dimethyl sulfoxide or PPT (100 nM) for 12 h. Luciferase assays were performed according to manufacturer's instructions (Promega, Dual-Glo Stop & Glo). Data are expressed as a ratio of luciferase to Renilla relative light units and converted to fold-change from basal.
Chromatin Immunoprecipitation (ChIP)-Stable ER␣-expressing 3T3-L1 cells (described above) were used for ChIP experiments. Cells were grown to confluence and at day 3 of differentiation were harvested and subjected to ChIP analysis as previously described (31) using antibodies against ER␣ (Santa Cruz Biotechnology). Primers for detection of the presence of the Lcn2 promoter by qPCR are outlined in supplemental Table S4.
Immunoblot Analysis-Animal tissues and cultured cells were homogenized in RIPA lysis buffer containing protease and phosphatase inhibitors. Samples were separated by SDS-PAGE, transferred to PVDF membrane, and probed with the following antibodies for protein detection: pan-actin (Cell Signaling Technologies), ER␣ (Santa Cruz Biotechnology), and Lcn2 (goat anti-mouse, R&D Systems). Densitometric analysis was performed using Bio-Rad Chemidoc image software, Quantity One (version 4.6).
Adipose Tissue Histology-Periovarian adipose tissue immunohistochemistry was performed on paraformaldehyde (8%) fixed samples by the University of California, San Diego, Moores Cancer Center, Histology and Immunohistochemistry Shared Resource Facility, as previously described (32). Adipocyte size was estimated from H&E-stained sections of adipose tissue by analyzing 10 different sections per genotype, counting 100 adipocytes per section using a Nikon Eclipse 80i Upright microscope and Image-Pro Plus Software (Media Cybernetics).
Statistical Analyses-Associations between biomarkers in human subjects were analyzed using Spearman's rho correlation. Relationships between expression levels and % body fat for mice were determined by Pearson's correlation coefficient. Differences between subgroups (normal, underweight, overweight, and obese) were compared using the Student's t test. Multiple group by treatment mean comparisons were performed using ANOVA with Tukey's LSD post-hoc test analyses. Significance was established a priori p Ͻ 0.05.

Adipose-specific Deletion of ER␣ Promotes Obesity-Previous
studies have shown that the expression of ESR1 is significantly reduced in adipose tissue from obese women (16). Indeed, we observed reductions in Esr1 expression in adipose tissue from genetically obese Lep Ob male and female mice (Fig. 1A). We also found that adipose tissue Esr1 expression is inversely correlated with the degree of HFD-induced adiposity in more than 100 genetically diverse strains of inbred female mice, termed the UCLA HMDP (23) (Fig. 1B).
To determine whether impaired ER␣ action promotes adiposity even during normal chow feeding, we generated fat-specific ER␣ knock-out mice (FERKO) by mating floxed ER␣ (f/f) animals (33) with transgenic mice expressing Cre recombinase driven by the aP2 promoter (20). As expected, Esr1 expression levels were significantly reduced in both white adipose tissue (WAT) and brown adipose tissue (BAT) of female (Fig. 1C) and male (Fig. 1D) FERKO mice, compared with f/f Control littermates. Deletion was selective for adipose tissue as Esr1 levels were maintained in skeletal muscle and liver of female FERKO mice (Fig. 1, C and D). Phenotypic analyses revealed that the body mass of female FERKO mice was increased compared with that of f/f Control animals (26.3 Ϯ 0.14 versus 23.0 Ϯ 0.13 g, p Ͻ 0.001), a finding also observed in male mice, although the difference did not reach statistical significance (p ϭ 0.116) (Fig.  1E). Consistent with findings for body weight, FERKO female mice had significantly larger gonadal fat pads (0.43 Ϯ 0.01 versus 0.63 Ϯ 0.01 g, p ϭ 0.001; Fig. 1F), BAT depots (0.046 Ϯ 0.001 versus 0.067 Ϯ 0.001 g, p ϭ 0.029, Fig. 1G), and liver weight (1.32 Ϯ 0.02 versus 1.01 Ϯ 0.01 g, p ϭ 0.002; Fig. 1H) compared with f/f Control. Adipocytes from gonadal fat pads of female FERKO mice were larger in diameter than those from f/f Control mice (Fig. 1I). In contrast to females, WAT (p ϭ 0.03) but not BAT (p ϭ 0.1) and liver (p ϭ 0.4) weights were elevated for male FERKO versus f/f Control.
Analyses of circulating factors were consistent with the observed obesity phenotype in female FERKO mice as plasma levels of leptin (2.4 Ϯ 0.1 versus 1.2 Ϯ 0.1 ng/ml, p ϭ 0.04) and the inflammatory marker PAI-1 (1.0 Ϯ 0.01 versus 0.5 Ϯ 0.04 ng/ml, p ϭ 0.05) were increased over that of f/f Control animals ( Table 1). In contrast, no differences in plasma concentrations of insulin, adiponectin, or estradiol were detected between the genotypes (Table 1).

Adipose Tissue ER␣ Deletion Alters Adipocyte Function and Lipocalin 2 Expression in Female
Mice-Considering that the phenotype of female FERKO mice was more prominent than that observed for males, we focused subsequent screening studies on adipose tissue from female animals. Expression of the adipogenic transcription factors C/EBP␣ and poly(ADP-ribose) polymerase ␥ were reduced in FERKO fat ( Fig. 2A). With the exception of the inflammatory chemokine MCP1, which was elevated in FERKO ϳ5-fold above f/f Control (p ϭ 0.001; Fig. 2B), the expression levels of proinflammatory cytokines were similar between the genotypes of female mice. Analysis of adipokine expression from FERKO fat showed no change in adiponectin, leptin, or Rbp4 transcript levels. In contrast, a significant 9-fold elevation in the expression of Lcn2 was observed in adipose tissue from FERKO compared with f/f Control (Fig.  2C). Increased Lcn2 expression was also detected in WAT from Lep Ob mice of both genders compared with the respective lean controls (Fig. 2D). Of note, a gender dimorphism in adipose tissue Lcn2 expression was also observed as transcript levels were markedly elevated in adipose tissue of lean WT and f/f Control male mice compared with similar genotype female animals (Fig. 2, D and E). Our data suggest that this gender dimorphism was a possible result of higher basal Esr1 expression levels in WAT of female versus male mice (Fig. 1A). Consistent with this observed gender dimorphism, in male Lep Ob mice an experimental increase in circulating estradiol concentration to levels observed in normally cycling females reduced expression of Lcn2 in epididymal fat below that observed for lean male animals (Fig. 2D).
In gonadal white and subscapular brown adipose tissue, the loss of ER␣ promoted increased expression of Lcn2 (5-10-fold) in FERKO compared with f/f Control for female but not male animals (Fig. 2, E and F). Similar to periovarian adipose, reduced Esr1 expression in subcutaneous fat from FERKO female mice (Fig. 2G) was paralleled by a marked increase in Lcn2 expression over f/f Control (Fig. 2H). However, we

Adipose Tissue ER␣ Expression, Obesity, and Breast Cancer Progression
detected increased Lcn2 expression levels in SC over gonadal fat from all female animals studied. Immunoblot analyses confirmed that the Lcn2 protein was significantly elevated in periovarian WAT from female FERKO mice versus f/f Control (Fig.  2I). Similar to findings for transcript abundance, Lcn2 protein levels in WAT from male mice were significantly elevated above levels for females. Moreover, no difference in WAT Lcn2 protein was detected between the genotypes in male animals (Fig. 2I). In line with reports that circulating Lcn2 levels are elevated in the plasma of obese, insulin-resistant mice and humans (34 -37), we found that plasma Lcn2 levels were also elevated in female FERKO compared with f/f Control mice (402 Ϯ 7.5 versus 361 Ϯ 1.4 pg/ml, respectively, p ϭ 0.02; Fig.  2J). In addition to showing that ER␣ deletion increased Lcn2  FEBRUARY 27, 2015 • VOLUME 290 • NUMBER 9 expression in adipose tissue of female mice, we confirmed a strong inverse relationship between Esr1 and Lcn2 expression in WAT obtained from 102 unique strains of HF/HS-fed female mice of the UCLA HMDP (Fig. 2K). Lcn2 expression was highly correlated with body fat percentage (Fig. 2L), similar to the relationship between adiposity and the adipokine leptin. These data confirm that natural genetic variation of Esr1 expression in adipose tissue among diverse mouse strains is inversely associated with Lcn2 expression, an adipokine we found tightly correlated with adiposity in female mice.

ER␣ Deletion Impairs Adipocyte Function and Increases Lcn2
Expression in Culture-Prior studies have identified a functional estrogen response element (ERE) within the Lcn2 gene promoter that controls its expression in normal breast tissue (38). We investigated the potential for ER␣ to directly modulate Lcn2 expression in adipocytes. Our own computational analysis confirmed the existence of a complete palindromic ERE at Ϫ2533 bp upstream from the transcriptional start site of the mouse Lcn2 gene (38). To demonstrate direct action of ER␣ on Lcn2 expression independent of secondary in vivo cues including insulin resistance or obesity, we showed that Lcn2 expression was elevated in cultured primary adipocytes from FERKO WAT compared with primary cells isolated from WAT of f/f Control animals (Fig. 3A). Primary adipocytes lacking ER␣ (FERKO) released elevated amounts of Lcn2 into the medium (Day 3 of differentiation) compared with control ER␣ replete adipocytes (Fig. 3B).
Next we studied the effects of 17 ␤-estradiol (E 2 ) on Lcn2 expression in 3T3-L1 adipocytes. Maximal suppression of Lcn2 expression was achieved with 1 nM E 2 treatment (Fig. 3C). To further support a role for ER␣ in the regulation of Lcn2, we stably overexpressed ER␣ using retrovirus. Lcn2 expression was reduced by ϳ80% in cells overexpressing ER␣ compared with vector controls (Fig. 3D). To investigate the effects of ER␣ lossof-function on Lcn2 expression, we generated four 3T3-L1 ER␣ knockdown (Esr1-KD) cell lines (Esr1-KD B, C, D, and E) by stably expressing Esr1 shRNAs (Fig. 3E). Three of the cell lines (C, D, and E) showed significantly reduced Esr1 expression, with a concomitant increase in expression of the Lcn2 transcript (Fig. 3E). Of note, the KD cell line that showed no reduction in Esr1 (Esr1-KD-B), also showed no change in Lcn2 expression (Fig. 3E). Because of its robust action, Esr1-KD D was selected for subsequent analyses. Similar to our observations for primary adipocytes from FERKO mice, Esr1-KD 3T3-L1 adipocytes showed significantly reduced ER␣ protein expression (Fig. 3F) and increased secretion of Lcn2 into the culture media (Fig. 3G). No morphological defects were observed in Esr1-KD 3T3-L1 adipocytes, as confluent, non-differentiated pre-adipocytes (day 0, Fig. 4) or as non-lipid loaded differentiating cells (day 3). However, after 4 days of lipid loading in the presence of insulin (day 7, Fig. 4), Esr1-KD cells displayed a reduction in lipid accumulation compared with Scr-Control cells. Furthermore, differentiated Esr1-KD adipocytes were larger in size and possessed substan-tially larger intracellular lipid droplets than WT cells (Fig. 4,  arrows, bottom panel). These data are consistent with the observation of larger adipocytes and reduced expression of adipocyte transcription factors in FERKO adipose tissue.
ER␣ Binds the Lcn2 Promoter and Competes with C/EBP for Regulation of Lcn2 Expression-Considering that previous studies have shown Lcn2 expression is induced by the C/EBP family of transcription factors (34), and that ER␣ forms complexes with members of this family to control their action on specific target genes (39,40), we hypothesized that interaction of ER␣ with C/EBP transcription factors may be critical in the regulation of Lcn2 expression specifically in adipocytes. The observation that a C/EBP binding site (ccaat) overlaps with the ERE in the Lcn2 promoter (Ϫ2536 bp), supported our hypothesis and led us to investigate the role of ER␣ in modulating the function of C/EBP factors in the control of Lcn2 expression. To test our hypothesis further, we generated a luciferase reporter construct (pGL4) containing the proximal promoter of mouse Lcn2 (2.7 kb) including the ERE (Fig. 5A). Transient transfection of this luciferase reporter construct into 3T3-L1 adipocytes alone or together with ER␣ and C/EBP factor expression vectors showed that ER␣ suppressed C/EBP-induced luciferase activity (Fig. 5B). ChIP studies conducted in 3T3-L1 adipocytes confirmed ER␣ binding to the Lcn2 proximal promoter construct at two sites (Fig. 5C). To confirm the requirement of DNA binding for the repression of Lcn2 by ER␣ in vivo, we assessed Lcn2 expression in gonadal WAT from the ER␣ DNA binding incompetent mouse in which ER␣ protein tethering activity is maintained (21). In gonadal adipose tissue from these DNA binding domain mutant mice (ER␣-DBD⌬) we observed a marked increase in C/ebp␤ and Lcn2 expression levels over WT control (Fig. 5D). These data confirm the insufficiency of ER␣ protein tethering, but confirm the requirement of ER␣ DNA binding in the repression of Lcn2.
Lcn2 Promotes Proliferation and Migration of BrCA Cells-Considering that the phenotype of malignant cells can be influenced by the surrounding stromal microenvironment, and that adipocytes constitute the largest cellular compartment of breast tissue (41), we tested whether Lcn2, an adipocyte-secreted factor, could influence the behavior of BrCA cell lines. We treated SUM159PT BrCA cells with CM from adipocytes where Lcn2 was elevated in response to ER␣ knockdown, or CM from HEK293A cells transfected with an Lcn2 expression plasmid. Conditioned media from Esr1-KD adipocytes increased the proliferation of the breast carcinoma cell line SUM159PT compared with CM from Scr-Control adipocytes (Fig. 6A). Similarly, CM containing Lcn2 expressed from trans-fected HEK293A cells promoted increased SUM159PT cell proliferation compared with CM from cells transfected with a GFP expression plasmid (Fig. 6B), suggesting a direct effect of Lcn2 on the proliferation of SUM159PT BrCA cells. Additionally, both Lcn2-CM and Esr1-KD CM increased the migration of SUM159PT cells over scratched monolayers, but had no effect on immortalized, non-transformed breast epithelial, MCF10A cells (Fig. 6C).
To test the role of Lcn2 in mediating the proliferative and promigratory effects of CM from ER␣-deficient adipocytes on breast cancer cells, we deleted Lcn2 from ER␣-deficient adipocytes (Fig. 7, A and B). Findings for proliferation and migration of SUM159PT BrCA cells treated with CM from adipocytes with a dual KD of Esr1 and Lcn2 mirrored findings for cells treated with CM from Scr-Control ER␣ replete adipocytes (Fig. 7, C and D). These data indicate that increased Lcn2 released from ER␣-deficient adipocytes is a critical mediator of BrCA cell proliferation and migration.
Expression of Lcn2 Signaling Components in BrCA Cells Confer Lcn2 Responsiveness-In an effort to more rigorously interrogate the effect of Lcn2 on BrCA cells in vitro, we studied the Lcn2 responsiveness of additional BrCA cell lines (Fig. 8). We observed a wide variability in the proliferation response to Lcn2 in the different BrCA cell lines (ZR75-1, MCF7, T47D, and EFM19), as well as between lines arising from similar cellular origins. The advantage of studying these specific cell lines is that each is cultured using the same media and environment thus eliminating confounding factors arising from varying culture conditions (27). Lcn2-stimulated proliferation was highest in MCF7 and EFM19 and negligible in ZR75-1 cells (data not shown). Interestingly, cellular proliferation correlated well with the expression of BDH2, the intracellular rate-limiting enzyme that catalyzes the production of the mammalian siderophore 2,5-dihydroxybenzoic acid necessary for iron homeostasis and overall Lcn2 responsiveness (42) (Fig. 8A).
In contrast to findings showing that LCN2 was expressed in a number of human cancers and associated with ER/progesterone receptor status (43), we screened 55 tumor samples as well as 53 cancer cell lines and found no correlation between ESR1 and LCN2 expression or LCN2 expression and tumor type (supplemental Table 2; cell lines as reviewed in Ref. 27). LCN2 expression levels were ϳ100-fold higher in ZR75-1 cells than MCF7 cells, and ϳ1000-fold higher than EFM19 cells (Fig. 8B). These findings raise an important question for future study regarding the relative roles of intracellular versus extracellular Lcn2 in driving neoplastic development and metastatic potential.
Differences in components of the Lcn2 signaling pathway were also apparent between lines, with elevated expression of the Lcn2 receptor, SLC22A17 (variant 1), found in the most responsive cell types (MCF7 and EFM19) (Fig. 8C). Exogenous Lcn2 was found to only modulate its receptor expression in MCF7 cells, whereas all other BrCA cell types were unresponsive (Fig. 8D). Expression levels of ferritin light chain (FLT), BNIP3, SNAIL, SLUG, and Vimentin, markers of EMT and apoptosis susceptibility, were elevated in cell lines where ectopic Lcn2 promoted increased proliferation (Fig. 8, E-I). Collectively, these data suggest that intrinsic differences in expression of components of the Lcn2 signaling pathway may underlie the differential responses of BrCA cells to the paracrine actions of Lcn2.
BDH2 Expression in Human BC Tumors Correlates with BMI and Lcn2, but Not Leptin or Adiponectin-To investigate a potential role for components of the Lcn2 signaling pathway in the differential response of primary BrCA tumors to obesity, we analyzed data from microarray studies performed on biopsied tumor samples from women of varying BMI participating in the TORI B02 study (24). Plasma was collected at the time of diagnosis, prior to therapeutic intervention. As expected, plasma adipokines, leptin, and adiponectin were significantly correlated with BMI. Similar to published findings showing Values are expressed as mean Ϯ S.E., and mean differences were detected by Student's t test and ANOVA where appropriate. For each analysis, three independent studies were performed in triplicate. *, p Ͻ 0.05 between cell type difference for a given treatment; #, Ͻ 0.05 between treatment difference for a given cell type. FEBRUARY 27, 2015 • VOLUME 290 • NUMBER 9 JOURNAL OF BIOLOGICAL CHEMISTRY 5575 increased plasma Lcn2 levels in obese subjects (34,36,37), the mean plasma Lcn2 concentration was elevated 23% in patients with a BMI Ͼ30 (Fig. 9A). Additionally, we identified a significant positive association between tumor expression of BDH2 and BMI, as well as tumor BDH2 expression and plasma Lcn2 concentration (Fig. 9, B and C). Unfortunately, probes for the Lcn2 receptor SLC22A17 were not present in the microarray platform so we were unable to assess the relationship between obesity and tumor Lcn2 receptor expression. In contrast to our findings for Lcn2, we found no significant association between tumor BDH2 expression and circulating levels of leptin or adiponectin (Fig. 9, D and E), indicating that tumor BDH2 expression is unlikely influenced by these circulating factors previously associated with BrCA progression (as reviewed in Refs. 44 -46). Moreover, no significant association between plasma levels of Lcn2 and adiponectin or leptin were detected (Fig. 9, F  and G). Thus, when considered in aggregate, these data suggest that adipocyte expression of ESR1, and not simply adiposity per se, may play a greater role in controlling Lcn2 production and breast cancer risk.

DISCUSSION
Obesity is an established risk factor for chronic diseases including atherosclerosis, type 2 diabetes, and cancers of the liver, colon, and breast. Normally cycling females are partially protected against these diseases, but much of this protection is lost following menopause when increased adipose tissue weight gain, tissue inflammation, and metabolic dysfunction manifests (47). Unfortunately, the mechanistic underpinnings linking obesity and metabolic dysfunction to cancer pathobiology remain incompletely understood. Studies by Nilsson et al. (16) implicated a relationship between reduced expression of ER␣ in adipose tissue and obesity in women. However, whether reduced adipose tissue ESR1 expression is causal for obesity and a disease-promoting secretory profile was unknown.
Herein, we reveal a mechanism whereby genetic ablation of Esr1 in adipose tissue of female mice promotes increased adi-posity, tissue inflammation, and increased expression and secretion of the adipokine Lcn2. Considering that the phenotype of malignant cells can be influenced by surrounding stromal cells comprising the tumor microenvironment, and that adipocytes constitute the largest cellular compartment of breast tissue (41), we tested whether adipocyte-secreted factors could influence the behavior of BrCA cells. We showed that adipocytes lacking ER␣ secrete factors that exert marked effects on BrCA cell proliferation and migration. Because Lcn2 was highly expressed and secreted from ER␣-deficient adipocytes, we next determined whether Lcn2 was a critical factor driving the proliferative and migratory response of BrCA cells. Consistent with our hypothesis, deletion of Lcn2 from ER␣-deficient adipocytes reversed the proliferative and pro-migratory effects of CM on SUM159PT BrCA cells. These findings suggest that Lcn2 is an important potential mediator promoting tumorigen-  FEBRUARY 27, 2015 • VOLUME 290 • NUMBER 9 esis in the context of obesity. Extending this link clinically in BrCA patients, we found a significant correlation between obesity and tumor expression of BDH2, an enzyme critical for cellular responsiveness to Lcn2 (48). These studies describe a previously unrecognized mechanism that connects obesity and breast cancer through ER␣ regulation of Lcn2 in adipocytes.

Adipose Tissue ER␣ Expression, Obesity, and Breast Cancer Progression
Although prior studies have demonstrated a role for Lcn2 in the proliferation and migration of cancer cells in culture (35,49,50) and have shown that LCN2 is strongly associated with breast cancer stage in microarray and expression profiling studies, the exact cellular source or sources of Lcn2 have not yet been determined (35,43,51). Additionally, although effects of Lcn2 on tumor burden in mouse models of breast cancer have been observed (52)(53)(54), these studies in murine models have failed to identify the mechanisms by which Lcn2 promotes tumorigenesis in vivo. Because Lcn2 is now considered a clinical predictor of disease prognosis in human primary breast cancer patients (43,55), resolution of the mechanisms regulating LCN2 expression, the identification of the cell types involved in its paracrine and endocrine actions, as well as determination of the cell and tissue-specific effects of Lcn2 signal transduction, require greater attention. Our data strongly support a link between ER␣-mediated regulation of Lcn2 in adipose tissue and breast cancer progression; however, our studies are limited to the direct impact of Lcn2 on breast cancer cell behavior and do not extend to the potential effects of Lcn2 on the tumor microenvironment, which is likely of equal or greater importance for disease prognosis.
Previous work has shown that Lcn2 expression is up-regulated in differentiating adipocytes in culture, and that circulating levels of Lcn2 are associated with obesity and metabolic dysfunction in wild type male mice (34). Our findings presented herein for female mice now reproduce those initial studies conducted in males. Additionally we have identified ER␣ as a critical transcriptional regulator of Lcn2 production and adiposity.
In line with the notion that obesity and alteration in adipocyte behavior underlie tumor progression, it is shown that mature breast adipocytes, but not pre-adipocytes, promote breast cancer cell motility (56). Thus it follows that the secretion of soluble factors from differentiated adipocytes may pro- FIGURE 9. Obesity is associated with increased BDH2 expression in primary tumors from breast cancer patients. A, plasma Lcn2 is elevated in obese (BMI Ͼ 30; n ϭ 26) compared with lean (BMI Ͻ 25; n ϭ 31) women diagnosed with BrCA. BDH2 expression is significantly associated with BMI (B) and plasma (C) Lcn2 concentration in female breast cancer patients from the TORI trial. No significant relationship between BDH2 expression and circulating levels of Lcn2 versus leptin (D and F) or the anti-inflammatory adipokine adiponectin (E and G) were detected. Values are expressed as mean Ϯ S.E., and mean differences were detected by Student's t test. *, p Ͻ 0.05 between groups. Correlation coefficients were determined by Pearson's r, and p values are provided for each correlation. vide critical cellular cross-talk necessary for disease progression (9). Interestingly, cancer-associated adipocytes located at the tumor invasive front appear delipidated and resemble a de-differentiated phenotype compared with more distal adipocytes. These front-line adipocytes are characterized by a fibroblastlike morphology and down-regulation of key adipogenic markers including poly(ADP-ribose) polymerase ␥, C/EBP␣, and aP2 (9), a finding reproduced in ER␣-deficient adipocytes. These observations suggest that cancer cells may engage in reciprocal talk altering adipocyte phenotype to promote mobilization of substrate, extracellular matrix remodeling, and angiogenesis (41). Scherer and colleagues (57) previously showed that adipocyte-secreted factors have an unparalleled ability to promote increased cell motility, migration, and tumor angiogenesis compared with secreted factors of other stromal cell types. In aggregate, our studies suggest these cancer-promoting effects of adipocytes are accentuated by ER␣ deficiency.
Although the physiological role of Lcn2 has remained incompletely understood since its initial discovery, it has been primarily associated with iron metabolism and metalloproteinase (MMP-9 specifically) activity (58). More recently, Green and colleagues (28,48) have identified and characterized both the receptor for Lcn2 (Slc22a17) and enzymes involved in intracellular signaling including Bdh2. We found that BDH2, an enzyme that catalyzes the formation of the siderophore that binds Lcn2 to control cellular iron metabolism (48), was highly correlated with BrCA cell proliferation. Considering that circulating concentrations of Lcn2 and breast tumor BDH2 expression levels were elevated in obese subjects and that RNAi-induced BDH2 deficiency promotes BrCA cell apoptosis susceptibility (48), suggests that targeting this pathway to restrain or reverse breast tumor development in the context of obesity may be of therapeutic benefit.
Additionally, given that adjuvant breast cancer therapies aimed at reducing estrogen production and antagonizing ER␣ action in mammary tumors also exert whole body effects, the long-term impact of these therapeutic strategies on metabolic function should be monitored, especially in women susceptible for obesity and type 2 diabetes. Moreover, because obesity and type 2 diabetes reduce overall breast cancer survival rates in part by diminishing effectiveness of conventional anti-tumor therapeutics (59,60), novel approaches to restrain breast cancer progression and reduce mortality in obese populations should be pursued.
In summary, the current investigation provides evidence that loss of ER␣ signaling in adipose tissue promotes obesity and induces the expression of Lcn2, a glycoprotein implicated in metabolic dysfunction and breast tumorigenesis and metastasis. Our findings suggest that reduced ER␣ action in adipose tissue, such as that associated with natural genetic variation, menopause, or obesity may be mechanistically linked to the increased prevalence of breast cancer observed under these conditions. Considering that obesity accounts for ϳ20% of all cancer deaths in women over age 50 (61), and based upon our findings that ER␣ deletion in adipose is causal of obesity, strategies to maintain ER␣ action in metabolic tissues including adipose tissue may be of benefit for cancer prevention.