Identification of Extracellular and Intracellular Signaling Components of the Mammary Adipose Tissue and Its Interstitial Fluid in High Risk Breast Cancer Patients

It has become clear that growth and progression of breast tumor cells not only depend on their malignant potential but also on factors present in the tumor microenvironment. Of the cell types that constitute the mammary stroma, the adipocytes are perhaps the least well studied despite the fact that they represent one of the most prominent cell types surrounding the breast tumor cells. There is compelling evidence demonstrating a role for the mammary fat pad in mammary gland development, and some studies have revealed the ability of fat tissue to augment the growth and ability to metastasize of mammary carcinoma cells. Very little is known, however, about which factors adipocytes produce that may orchestrate these actions and how this may come about. In an effort to shed some light on these questions, we present here a detailed proteomic analysis, using two-dimensional gel-based technology, mass spectrometry, immunoblotting, and antibody arrays, of adipose cells and interstitial fluid of fresh fat tissue samples collected from sites topologically distant from the tumors of high risk breast cancer patients that underwent mastectomy and that were not treated prior to surgery. A total of 359 unique proteins were identified, including numerous signaling molecules, hormones, cytokines, and growth factors, involved in a variety of biological processes such as signal transduction and cell communication; energy metabolism; protein metabolism; cell growth and/or maintenance; immune response; transport; regulation of nucleobase, nucleoside, and nucleic acid metabolism; and apoptosis. Apart from providing a comprehensive overview of the mammary fat proteome and its interstitial fluid, the results offer some insight as to the role of adipocytes in the breast tumor microenvironment and provide a first glance of their molecular cellular circuitry. In addition, the results open new possibilities to the study of obesity, which has a strong association with type 2 diabetes, hypertension, and coronary heart disease.

It has become clear that growth and progression of breast tumor cells not only depend on their malignant potential but also on factors present in the tumor microenvironment. Of the cell types that constitute the mammary stroma, the adipocytes are perhaps the least well studied despite the fact that they represent one of the most prominent cell types surrounding the breast tumor cells. There is compelling evidence demonstrating a role for the mammary fat pad in mammary gland development, and some studies have revealed the ability of fat tissue to augment the growth and ability to metastasize of mammary carcinoma cells. Very little is known, however, about which factors adipocytes produce that may orchestrate these actions and how this may come about. In an effort to shed some light on these questions, we present here a detailed proteomic analysis, using two-dimensional gel-based technology, mass spectrometry, immunoblotting, and antibody arrays, of adipose cells and interstitial fluid of fresh fat tissue samples collected from sites topologically distant from the tumors of high risk breast cancer patients that underwent mastectomy and that were not treated prior to surgery. A total of 359 unique proteins were identified, including numerous signaling molecules, hormones, cytokines, and growth factors, involved in a variety of biological processes such as signal transduction and cell communication; energy metabolism; protein metabolism; cell growth and/or maintenance; immune response; transport; regulation of nucleobase, nucleoside, and nucleic acid metabolism; and apoptosis. Apart from providing a comprehensive overview of the mammary fat proteome and its interstitial fluid, the results offer some insight as to the role of adipocytes in the breast tumor microenvironment and provide a first glance of their molecular cellular circuitry. In addition, the results open new possibilities to the study of obesity, which has a strong association with type 2 diabetes, hypertension, and coronary heart disease.

Molecular & Cellular Proteomics 4: 492-522, 2005.
During the last years there have been numerous reports indicating that growth and progression of breast as well as other tumor cells depend not only on their malignant potential but also on stromal factors present in the tumor microenvironment, the insoluble extracellular matrix as well as cell-cell interactions (Refs. 1-6 and references therein). Of all the cell types present in the microenvironment, which include endothelial cells and their supporting pericytes, inflammatory cells (neutrophils, macrophages, eosinophils, and mast cells), immune cells (lymphocytes and dendritic cells), smooth muscle cells, myofibroblasts, preadipocytes, and adipocytes, the last are perhaps the least well studied despite the fact that they correspond to one of the most prominent cell types surrounding the breast tumor cells (7).
Until recently, adipocytes were mainly considered as an energy storage depot, but we now have clear evidence pointing to the fat tissue as an endocrine organ that produces hormones, growth factors, adipokines, and other molecules that may affect normal duct development as well as tumor growth and metastasis (Refs. 8 -18 and references therein). It has been shown that normal mouse mammary ducts do not form correctly if there is no proper interaction with the fat tissue, and a number of signaling pathways that may be involved in this process have been identified (8 -10). Elliot and colleagues (19), on the other hand, showed that fat tissue is able to augment the growth and ability to metastasize of the murine mammary carcinoma cell line SP1 when injected subcutaneously or peritoneally far away from fat pads, and Iyengar and colleagues (7) reported that factors secreted by adipocytes promote mammary tumorigenesis through induction of antiapoptotic transcriptional programs and proto-oncogene stabilization. In addition, several reports have demonstrated an association between breast cancer growth and the presence of adipose tissue (20 -22), and a connection between obesity and increased incidence of cancer has been established for breast, colorectal, endometrial, renal (renal cell), and esophageal (adenocarcinoma) malignancies (Ref. 23 and references therein). Presently, however, there is only limited information as to the factors produced by adipocytes that may affect normal breast duct development and tumor progression (1,7,24).
Most studies of adipogenesis have made use of rodent cells or primary cultures of human mesenchymal stem cells that have been induced to differentiate into adipocytes using a variety of effectors. By using cDNA microarrays (25)(26)(27)(28)(29)(30)(31)(32) and proteomic technologies (Refs. 24 and 33 and references therein), it has been possible to identify several genes and proteins that are differentially regulated as a result of adipogenesis. These studies have been inspired by the facts that increased adiposity and a failure in adipocyte differentiation are associated with morbidity, mortality, and many disorders, including obesity, which has a strong association with type 2 diabetes (34,35), hypertension, and coronary heart disease (36). The question remains, however, as to whether these experimental model systems are able to completely replicate the in vivo situation (Ref. 37 and references therein) as it has been shown that gene expression changes associated with adipogenesis in vivo and in vitro, while sharing many features in common, are in some respects rather different (32).
In vivo transcript profiling studies of human and murine fat tissue (28,32,38) have shown the complexity of the adipocyte transcriptome and have implicitly established that the biology of these cells has a degree of intricacy that was not expected. To date, however, only two studies have investigated the proteome of human and mouse adipose tissue; one resolved about 100 human proteins using wide IPG strips and identified 16 polypeptides by means of mass spectrometry (39), while the other resolved a considerable number of murine white adipose tissue proteins and identified 80 unique polypeptides (40).
In our translational breast cancer program, which involves high risk patients that have undergone mastectomy (41)(42)(43), we have frequently detected tumor cells interdigitating with and spreading through the peripheral fat tissue suggesting a close association between these two cell types ( Fig. 1) (7). This observation together with published data demonstrating a role for the fat tissue in mammary gland development (Refs. 1 and 44 and references therein) and in modulating tumor behavior (3,(45)(46)(47) prompted us to carry out a detailed proteomic analysis of fresh fat tissue and its interstitial fluid in an attempt to identify protein components and excreted factors that may shed some light on the close association between mammary epithelia and fat tissue. In the first instance and to simplify the study, we chose to analyze fat tissue located topologically distant from the tumor in high risk breast cancer patients registered at the Department of Breast and Endocrine Surgery, Copenhagen University Hospital.

Sample Collection and Handling
Fat tissue biopsies from sites topologically distant from the tumor (more than 4 -5 cm away; Fig. 2A) of high risk patients 1 that underwent mastectomy were collected from the Pathology Department at the Copenhagen University Hospital 30 -45 min after surgery. Samples for gel analysis were placed in liquid nitrogen and were rapidly transported to the Institute of Cancer Biology where they were stored at Ϫ80°C. Samples for fluid recovery were placed in PBS, transported on ice, and processed immediately upon arrival at the Institute. On average, a total of 45 min to 1 h elapsed between surgical sample acquisition and sample preparation. The project was approved by the Scientific and Ethical Committee of the Copenhagen and Frederiksberg Municipalities (KF 01-069/03).

Fat Interstitial Fluid
About 0.50 g of clean fresh fat tissue was cut to small pieces (3 mm 3 ), dipped into 5 ml of PBS, and placed in a 10-ml conical plastic tube containing 1.0 ml of PBS. Samples were incubated for 1 h at 37°C in a humidified CO 2 incubator. Thereafter the samples were centrifuged at 1000 rpm for 2 min, and the supernatant was aspirated with the aid of an elongated Pasteur pipette. Samples were further centrifuged at 5000 rpm for 20 min in a refrigerated centrifuge (4°C). A fraction of the fat interstitial fluid (FIF) 2 (0.8 ml) was kept at Ϫ20°C for antibody array-based analysis, while the rest was freeze-dried and resuspended in 0.5 ml of O'Farrell lysis solution and kept at Ϫ20°C until use (48).

Two-dimensional Gel Electrophoresis and Western Immunoblotting
Twenty to thirty 6-m cryostat sections of frozen fat tissue were resuspended in 0.1 ml of CBL1 lysis solution (Zeptosens AG, Witterswil, Switzerland) and were kept at Ϫ20°C until use. In a few cases the sections were resuspended in 0.1 ml of O'Farrell lysis solution (48) with similar results. The advantages of the CBL1 lysis solution are that it yields better focused spots, and it does not dry so easily after prolonged storage. A detailed protocol of its use in the study of tumor tissue biopsies and cell lines will be the subject of a further publication. 3 Fat lysates and freeze-dried fluids resuspended in lysis solution were subjected to both IEF and NEPHGE two-dimensional (2D) PAGE as described previously (49). Between 30 and 40 l of sample were applied to the first dimension. Proteins were visualized using a silver staining procedure compatible with mass spectrometry analysis (50). Gels were dried between two pieces of cellophane. Western immunoblotting was performed as described previously (51).

Protein Identification by Mass Spectrometry
In-gel Digestion Protocol-Protein bands were excised from the dry gels followed by rehydration in water for 30 min at room temperature. The gel pieces were detached from the cellophane film, rinsed twice with water, and cut into about 1-mm 2 pieces with subsequent additional washes. Proteins were "in-gel" digested with bovine trypsin (unmodified, sequencing grade; Roche Diagnostics) for 8 h as described by Shevchenko and colleagues (52). The reaction was stopped by adding TFA (up to 0.4%) followed by shaking for 20 min at room temperature to increase peptide recovery. In most cases, peptides were analyzed using the supernatant. In the few cases where the amount of peptides was too low or when no conclusive identification was achieved by peptide fingerprinting using the supernatant, the remaining amount of supernatant (approximately 10 l) as well as the peptides additionally extracted from the gel pieces with 1% TFA and 50% ACN were concentrated on micro-ZipTip -C 18 columns in accordance with the manufacturer's protocol (Millipore). Peptides were eluted from the column with 50% ACN, 0.2% TFA directly on the target and co-crystallized with ␣-cyano matrix (2 mg/ml cyano-4-hydroxycinnamic acid in 0.5% TFA/ACN, 1:2, v/v). The extraction procedure strongly increased the amount of peptides, thus allowing direct sequence analysis of low intensity peptides.
Probe Preparation and Acquisition of the MALDI-TOF Spectra-Samples were prepared for analysis by applying 0.8 l of digested supernatant or microcolumn-eluted material on the surface of a 400/ 384 AnchorChip target (Bruker Daltonik, GmbH) followed by cocrystallization with 0.3 l of ␣-cyano matrix. After drying, the droplets were washed twice with 0.5% TFA to remove contamination from the samples.
Mass spectrometry was performed using a Reflex IV MALDI-TOF mass spectrometer equipped with a Scout 384 ion source. All spectra were obtained in positive reflector mode with delayed extraction using an accelerating voltage of 28 kV. Each spectrum represented an average of 100 -200 laser shots, depending on the signal-to-noise ratio. The resulting mass spectra were internally calibrated by using the autodigested tryptic mass values (805.417/906.505/1153.574/ 1433.721/2163.057/2273.160) visible in all spectra. Calibrated spectra were processed by the Xmass 5.1.1 and BioTools 2.1 software packages (Bruker Daltonik, GmbH). All spectra were analyzed manually.
Spectra originating from parallel protein digestions were compared pairwise to discard common peaks derived either from trypsin autodigestion or from contamination with keratins. Only unique peptides present in the spectra were used in the first search. Data base searching was performed against a comprehensive non-redundant data base using MASCOT 1.8 software (53) without restriction on the protein molecular mass and taxonomy. Since proteins were recovered from gels, a number of fixed modifications (acrylamide-modified cysteine, i.e. propionamide/carbamidomethylation) as well as variable ones (methionine oxidation and protein NH 2 terminus acetylation) were included in the search parameters. The peptide tolerance did not exceed 50 ppm, and as a maximum only one trypsin missed cleavage was allowed. The protein identifications were considered to be confident when the protein score of the hit exceeded the threshold significance score of 70 (p Ͻ 0.05) and not less than six peptides were recognized. The data base was checked for redundancy, and whenever it was possible the Swiss-Prot accession numbers were assigned. Additionally peptide mass fingerprinting analysis was performed using the MS-Fit program (Protein Prospector, University of California San Francisco Mass Spectrometry Facility, London, UK). We also used the Find-Mod software (Protein Prospector, University of California San Francisco Mass Spectrometry Facility, London, UK) to check any unmatched peptides for potential protein post-translational modifications. The second search was performed for all identifications as follows. 1) The predicted peptide digest was compared with the experimental one to reveal additional peptides present within the spectra; 2) the unmatched molecular weight values from the initial search were applied for extra search with the same reproducibility requirements for identification of the second and the third components in the spot. Whenever the protein score hit was close to the threshold significance score of 70, the PSD was performed as an additional means to confirm the identity of the proteins identified by post-translational modifications. The following PSD search parameters were used: peptide tolerance, 50 ppm; MS/MS tolerance, 1 Da without any restriction on the protein molecular mass and taxonomy. Since the amount of peptides extracted from the silver-stained gels did not yield overall peak intensities high enough to allow multiple peptide sequencing (prerequirement for conclusive PSD analysis), the identification of proteins was never made solely based on PSD analysis. The molecular weight and pI of the identified proteins were evaluated by analysis of mobility of the corresponding protein band in the 2D gel images. Positive protein identification was achieved in 80% of the cases with average sequence coverage of ϳ33%.

Preparation of Triton X-100 Extracts of Fat Tissue
Fresh fat tissue cut in small pieces (about 0.5 g) was homogenized with 3 ml of 0.1% Triton X-100 in PBS for 3 min at room temperature (54). Thereafter the samples were centrifuged at 1000 rpm for 2 min, and the supernatant was aspirated with the aid of an elongated Pasteur pipette. Samples were further centrifuged at 5000 rpm for 20 min in a refrigerated centrifuge (4°C). A fraction of the supernatant 2 The abbreviations used are: FIF, fat interstitial fluid; 2D, twodimensional; A-FABP, adipocyte fatty acid-binding protein; TIF, tumor interstitial fluid; IL, interleukin; TGF, transforming growth factor; TNF, tumor necrosis factor; IGF, insulin-like growth factor; CSF, colonystimulating factor; TIMP, tissue inhibitor of metalloproteinases; MAPK, mitogen-activated protein kinase; MDM2, mouse double minute 2; ER, estrogen receptor; ERK, extracellular signal-regulated kinase; CBP, cAMP-response element-binding protein (CREB)-binding protein; JNK, c-Jun NH 2 -terminal kinase; MEK, mitogen-activated protein kinase/extracellular signal-regulated kinase kinase. 3 P. Gromov, I. Gromova, and J. E. Celis, unpublished data.
was kept at Ϫ20°C for antibody array-based analysis, whereas the rest was freeze-dried and resuspended in lysis solution for 2D PAGE analysis (48,49).

Antibody Arrays for the Detection of Multiple Cytokines
Cytokines present in FIFs were detected using the RayBio® Human Cytokine Array C Series (RayBiotech, Inc.). Each array was incubated with 0.25 ml of FIF at 4°C overnight, and bound antigens were detected according to the manufacturer's instructions. The sensitivity of the antibodies present in the arrays ranges from 1-2000 pg/ml (for further details see www.raybiotech.com/human_array_sensitivity.pdf).

Antibody Array-based Identification of Key Cellular Effectors and Signaling Molecules
The relative level of cellular effectors and signaling molecules present in Triton X-100 fat tissue extracts were determined using the Panorama TM Ab Microarray-Cell Signaling array (Sigma). This array contains 224 different antibodies each spotted in two equal concentrations on nitrocellulose-coated glass slides. These antibodies represent several biological pathways including apoptosis, cell cycle, and signal transduction. Binding to a cognate antibody was detected by directly labeling the proteins in the cell extracts with a fluorescent dye according to the manufacturer's instructions.

Indirect Immunofluorescence
Fresh tumors containing marginal fat tissue were placed in formalin fixative and paraffin-embedded for archival use. Five-micrometer sections were cut from paraffin blocks of breast tumor and fat tissue, mounted on Super Frost Plus slides (Menzel-Glä ser, Braunschweig, Germany), baked at 60°C for 60 min, deparaffinized, and rehydrated   Table I. FIG. 4. NEPHGE 2D gel of whole fat tissue proteins stained with silver nitrate. Proteins indicated with a red arrow migrate both in NEPHGE and IEF gels and served as landmarks to align the gels. Selected proteins indicated with red arrows are expressed at low levels or are absent in the FIF (see also Fig. 9). A few proteins indicated with green arrows are enriched in the FIF (see also Fig. 9). The identity of the proteins indicated with numbers is given in Table I. through graded alcohol rinses. Heat-induced antigen retrieval was performed by immersing slides in 10 mM citrate buffer (pH 6.0) and microwaving in a 750-watt microwave oven for 10 min. The slides were then cooled at room temperature for 20 min and rinsed abundantly in tap water. Nonspecific staining of slides was blocked by incubation with 10% normal goat serum in PBS buffer for 30 min. Antigen was detected by overnight incubation at 4°C with a primary antibody at the appropriate dilution followed by a secondary antibody conjugated to Alexa Fluor® 488 or Alexa Fluor 594 (Molecular Probes, Eugene, OR). Sections were imaged using confocal laser scanning microscopy (Zeiss 510LSM). Proteins were identified using a combination of procedures that included mass spectrometry of proteins recovered from 2D gels ( Fig. 5 and Table I), 2D PAGE Western immunoblotting using specific antibodies ( Fig. 6 and Table I), and in a few cases by comparison with the master image of the keratinocyte 2D PAGE protein data base (proteomics.cancer.dk; Refs. 56 and 57). A complete list of all proteins identified in whole fat tissue extracts from high risk patients is given in Table I. Only in a few instances was it possible to confirm the presence of a given protein in the fat tissue using immunohistochemistry or immunofluorescence due to the high lipid content of the tissue. As an example, Fig. 2B shows a double immunofluorescence picture of a representative paraffin section from a fat tissue biopsy from a site distant to the tumor incubated with antibodies against A-FABP (fat cells, green) and the CD68 antigen (macrophages, red). No tumor cells were detected in this preparation as ascertained by staining with keratin 19specific antibodies (not shown).

Identification of Cellular Effectors and Signaling
Molecules in Triton X-100 Fat Extracts-Given the limitations inherent to 2D PAGE, we used multianalyte protein-based technologies to complement the gel-based proteomic analysis in an effort to detect lesser abundant cellular effectors and components of signaling pathways. To this end we treated fat tissue with Triton X-100 in PBS and incubated the extracts with the Panorama Ab Microarray-Cell Signaling array that contains                 224 different antibodies against components of various biological pathways (see Fig. 7) as described under "Experimental Procedures." Fig. 7 shows an array with the presence of 80 components detected in this particular case. A list of proteins identified by this approach is given in Table I. Protein Profiling of the FIF-FIF recovered from fresh fat tissue specimens dissected from sites topologically distant from the tumor of 20 high risk breast cancer patients ( Fig. 2A Table I. Low abundance cytokines and growth factors were detected by incubating FIF preparations with a multiple cytokine antibody array as described under "Experimental Procedures" (RayBio Human Cytokine Array C Series, RayBiotech, Inc.). As an illustrative example, Fig. 9 depicts arrays showing the presence of 98 cytokines and growth factors in FIF 48. Similar analysis of FIFs recovered from the fat tissue of several patients (results not shown) indicated that the cytokine patterns are quite similar among each other, albeit with some changes in their levels, but distinctive from the tumor interstitial fluid (compare FIF 48 with TIF 48; Fig. 9). Two proteins (granulocyte-macrophage colony-stimulating factor (CSF) and granulocyte CSF) were identified using the Bio-Rad Bioplex cytokine system as described previously (43). A complete list of proteins detected in the FIF, including cytokines and growth factors, is given in Table I.
Functional Classification-All in all, the study identified 359 primary translation products present in the fat tissue and its interstitial fluid (Table I). The molecular functions of these proteins as well as the biological process in which they participate were assigned in accordance with the Human Protein Resource Database (www.hprd.org) and are given in Table I. The list includes, but is not limited to, polypeptides involved in various biological processes such as signal transduction and cell communication (34%); energy metabolism (19%); protein metabolism (12%); cell growth and/or maintenance (10%); immune response (10%); transport (6%); regulation of nucleobase, nucleoside, and nucleic acid metabolism (5%); and apoptosis (3%). Approximately 1% of the proteins are of unknown function (Fig. 11). DISCUSSION A growing body of research indicates that tumor growth in vivo is dependent, in part, on adipose tissue. In addition, large epidemiology studies have found that obese women are at increased risk of developing postmenopausal breast cancer (58,59), and there is some evidence that environmental cues from adipocytes affect tumor cell survival, proliferation, differentiation, and migration. However, the exact mechanisms by which adipose tissue can promote an aggressive breast cancer phenotype are poorly understood. The results presented here provide for the first time a broad overview of the proteome of fresh mammary fat tissue and its interstitial fluid. The gel-based analysis resolved more than 1000 well focused proteins of which 183 unique polypeptides were identified using mass spectrometry (169 polypeptides), immunoblotting (7 polypeptides), and comparison with the keratinocyte protein database (7 polypeptides). Due to the inherent limitations of the gel-based approach, we extended these studies by using antibody arrays, which led to the identification of an additional 178 unique proteins, mainly signaling molecules, hormones, cytokines, and growth factors. Considering that one of the main functions of adipocytes is to act as an energy reservoir by storing lipids (37,60), it is not surprising that many of the proteins identified corresponded to components that play a role in energy metabolism ( Fig. 11 and Table I). These include many enzymes involved in lipid metabolism as well as the hormones leptin and adiponectin, both of which are components of the FIF. These hormones have been the subject of extensive research (17,(61)(62)(63)(64)(65)(66)(67)(68)(69)(70)(71)(72)(73)(74), and we will therefore restrict the discussion of our results to cytokines and growth factors secreted by adipocytes as well as to signaling pathways present in these cells.

Cytokines and Growth Factors Secreted by Adipocytes
There is compelling evidence indicating that adipocytes play a role in normal mammary epithelia development (1,44) as well as in tumorigenesis (1,7,19,22), and we have frequently detected tumor cells interdigitating with, and spreading through, the peripheral fat tissue in high risk breast cancer patients suggesting a close association between these cell types (Fig. 1). These observations prompted us to search for growth factors and cytokines that may be produced by adipocytes in vivo and that may lead to a better understanding of the mechanisms underlying this close association. To this end, we took advantage of a simple protocol that we devised for recovering the interstitial fluid that bathes the breast tumor microenvironment (TIF; Refs. 42 and 43). The TIF is composed of hundreds of proteins that are either secreted, shed by membrane vesicle-like exosomes (75)(76)(77)(78), and/or externalized due to cell death, and preliminary results indicated that the procedure could also be applied to fat tissue (42,43).
From our immunofluorescence analysis it seems likely that breast tumor cells and adipocytes provide mutual growth support to each other via the secretion of substances that may offer common benefits. For example, we often observed isolated preadipocytes in the stroma surrounding the breast tumor cells (Fig. 12, inset), suggesting that the latter secrete factors that directly, or indirectly through other cell types present in the microenvironment, commit pluripotent mesenchymal stem cells to adipocyte differentiation (85)(86)(87). Preadipocytes recruited in this way differentiate into mature adipocytes under suitable conditions and may provide the tumor cells with a friendly environment in which to spread (Fig. 12). Expanding adipose tissue requires active angiogenesis, and sites of neovascular angiogenesis are critical for tumor progression (88,89). The nature of the interplay between adipocytes and tumor cells is at present unknown, although the availability of the FIF may allow studying the effect of the adipokine mixture on breast epithelial cells using three-dimensional cultures of both normal and malignant breast tissue (10,41,44). Alternatively there are various model systems that have been described to study adipocyte differentiation (24,90), although these may not completely duplicate the in vivo situation (Refs. 32 and 37 and references therein).
It should be stressed that, in addition to the cytokines and growth factors mentioned above, the FIF also contains proteins that may have important functions in the tumor microenvironment. For example, we detected the presence of type VI collagen that has been described previously by Scherer and colleagues (12) in adipocyte tissue as an adipocyteenriched secretory protein that promotes pro-oncogenic pathways in breast cancer cells and that may play a key role in the regulation of normal and transformed mesenchymal cell proliferation in vitro (91) as well as in preventing apoptosis (92). It has also been shown that type VI collagen promotes the phosphorylation of glycogen synthase kinase-3␤ in malignant ductal epithelial cells, leading to increased ␤-catenin stabilization and transcriptional activity (7).

Signaling Pathways in Mammary Fat Tissue
By using Triton X-100 extracts of fat tissue in combination with multianalyte protein-based technologies for the detection of low abundance and phosphorylation-dependent key regulatory proteins it was possible to perform signal pathway profiling of mammary adipocytes focusing on discovery and quantitative analysis of proteins involved in a variety of bio- The identity of the proteins indicated with numbers is given in Table I. FIG. 9. NEPHGE 2D PAGE of FIF proteins stained with silver nitrate. Selected proteins indicated with red arrows are expressed at low levels or are absent in the FIF (see also Fig. 4). A few proteins indicated with green arrows are enriched in the FIF (see also Fig. 4). The identity of the proteins indicated with numbers is given in Table I. logical processes that include apoptosis, cell cycle, stress response, transcription, and signal transduction. The Panorama Ab Microarray-Cell Signaling array used for this purpose contains over 224 antibodies (for a description, see www. sigmaaldrich.com/img/assets/5941/Antibody_List.pdf) with some probes being specific for the phosphorylated protein of interest (e.g. focal adhesion kinase, MAPK, Raf, p38, Pyk2, p21-activated kinase 1, and death-associated protein kinase). In this way we were able to detect the presence of many low abundance signaling proteins in all biological processes examined, providing some insight into the cellular circuitry of adipocytes. We present below an overall functional view of the obtained results.
Cell Cycle-Adipose tissue is a dynamic body that changes in response to metabolic cues (93). Following key extracellular and intracellular signals, adipose tissue cell number increases through fat cell formation or adipogenesis, a process in which preadipocytes differentiate to become mitotically quiescent adipocytes (94). Adipocyte withdrawal from the cell cycle is presumably regulated by cell cycle inhibitors (95). We identified several cell cycle regulators, positive as well as negative (e.g. p53, MDM2, p21 Waf1 , p34 cdc2 , p19 INK4d , and Cdk4), and cyclins (cyclins B1, D2, and D3) reflecting the mixed nature of the samples, which contain both differentiated adipocytes as well as preadipocytes.
Apoptosis-Regulation of adipocyte cell number by apoptosis is thought to play an important role in adipose tissue homeostasis and to be part of a normal physiological cycle in adipocyte growth and development that can be altered under a variety of physiological and pathological conditions. Adipokines such as leptin, TNF-␣, and ciliary neurotrophic factor can induce adipose tissue apoptosis (96 -98), suggesting that adipose tissue cell number is regulated, at least in part, by an apoptotic signaling pathway involving caspase activation. Our cytokine array analysis showed that leptin, TNF-␣, and ciliary neurotrophic factor were all present in the samples examined ( Fig. 7 and Table I), and we identified several components of the caspase cascade including active effector caspase-3, and caspase-6, -7, -8, -10, and -11 as well as other apoptotic factors such as apoptosis-inducing factor, Smac/DIABLO, Bcl-x, and Bcl-10 ( Fig. 7 and Table I). We also identified components of the death receptor signaling pathway such as Fas and DAXX.
One factor we identified that is of great interest in the context of breast cancer is the estrogen receptor (ER) (Fig.  7 and Table I). Several findings indicate that mature human adipocytes possess ERs and thus might be an estrogenresponsive tissue (105,106). Recent work by Manabe and colleagues (21) has shown that mature adipocytes may be involved in the mechanisms regulating the growth of breast tumors through their growth-promoting effect on ERpositive tumor cells. Furthermore increased leptin levels in breast cancer patients are associated with enhanced blood plasma concentrations of progesterone and estradiol as well as enhanced tissue levels of ER and progesterone receptor suggesting that leptin stimulates the production of sex hormones (107). In view of these data and considering that breast cancer is a disease where treatment is largely based on antiestrogen therapy, the presence of large amounts of adipose tissue should be of major therapeutical concern and must be taken into consideration. Another class of transcription factors present in substantial amounts in the samples examined included histone acetylases (HAT1 and p300/CBP-associated factor) and deacetylases (histone deacetylases 1, 2, and 4), which most likely reflect a central role in the adipose cell differentiation program (108,109).

TGF-␤/Smad
Signaling-One of the most prominent features we observed in our array analysis was the relatively high level expression of Smad4 (Fig. 7), suggesting that TGF-␤ signaling (detected by cytokine array; Fig. 10) might play a key regulatory role in the mammary fat tissue of breast cancer patients given that repression of the activity of the key adipogenic transcription factors, CCAAT/enhancer-binding proteins, by Smad3/4 at CCAAT/enhancer-binding protein binding sites is known to block adipogenesis (110).
Nitric Oxide Signaling-Nitric oxide is involved in adipose tissue biology by influencing adipogenesis, insulin-stimulated glucose uptake, and lipolysis (111). We observed the presence of several enzymes responsible for nitric oxide formation (e.g. endothelial and inducible nitric-oxide synthases) in adipose cells (Fig. 7).
MAPK Signaling-Of the three MAPK pathways (p38MAPK, ERK1/2, and c-Jun NH 2 -terminal kinase (JNK)) we could identify components of two of them (p38MAPK and ERK1/2). However, although we could detect the presence of p38MAPK, we failed to detect the activated phosphorylated form of this protein, suggesting that p38 is present in mammary fat tissue in a latent form. This is consistent with the main known role that p38MAPK plays in adipose tissue metabolism as a central mediator of the cAMP signaling mechanism of brown fat that promotes thermogenesis by phosphorylating activating transcription factor 2 (Fig. 7, ATF2) (112). The presence of the phosphorylated form of the other MAPK pathway effector we observed, ERK1/2, as well as an epistatic component of the pathway (c-Raf) suggests that this pathway is active in mammary fat tissue. Signaling by the ERK pathway is reportedly involved in adipocyte cellular responses to cell size sensing (113), presence of adipokines (114,115), and differentiation (116,117). Consequently activity of this pathway in mammary fat tissue is consistent with a normal physiological cycle in adipocyte growth and development.
These data also suggest that ERK might be the most prominent of the MAPK signaling pathways in adipose tissue homeostasis in non-pathological conditions. Recent findings have shown that chronic activation of ERK, p38, or JNK can induce insulin resistance with the contribution of ERK being the strongest (118), lending some support to our observation.
Another important observation is the lack of JNK MAPK protein in the fat tissue samples. It has been shown that JNK is activated during obesity (119), and recent genetic and pharmacological data indicate that activated JNK could be critical in causing diabetes and insulin resistance (120). Thus, it would appear that the presence and subsequent activation of JNK in adipose tissue occurs only in response to adipose tissue metabolism-related stimuli, which would account for the lack of JNK expression in our samples.
Protein Kinase C and Phospholipase Signaling-Several reports demonstrated a functional role for the protein kinase B signaling pathway in adiposity (121,122). Consistent with these observations we found the presence of the inactive non-phosphorylated form of protein kinase B/AKT but not the phosphorylated forms (Thr 308 and Ser 473 ) in mammary fat tissue. We also identified several other proteins involved in phospholipase signaling such as MAPK phosphatase-1, PTEN, serum-and glucocorticoid-inducible kinase, and protein kinase D but not protein kinase C (␣, ␤, or ␥).
To conclude, our proteomic analysis is the most extensive carried out to date, and although the DNA microarray studies have identified many more genes, proteomics provided us with a glance at the gene products that are actually present in these cells. In particular, the large number of cytokines and growth factors secreted by adipocytes add to the complex mix of factors present in the fluid that bathes the tumor microenvironment (TIF) (45). Understanding how all these factors converge and regulate the social behavior of tumor cells represents a daunting scenario that may not be easy to recapitulate using current in vitro systems (32,123,124).
We would like to stress the fact that the large number of proteins present in the FIF (Table I) can have major implications for programs aiming at biomarker discovery in the blood because molecules secreted by adipocytes, and in the present study cytokines and growth factors in particular, add to the complex mixture of factors present in the tumor microenvironment, which is presumably the major source of molecules of predicted value that end up in the blood stream. Finally we would like to emphasize that the results presented here open new possibilities to the study of obesity and by association to type 2 diabetes (34,35), hypertension, and coronary heart disease (36).