Although high mammographic density is considered one of the strongest risk factors for invasive breast cancer, the genes involved in modulating this clinical feature are unknown. Tissues of high mammographic density share key histologic features with stromal components within malignant lesions of tumor tissues, specifically low adipocyte and high extracellular matrix (ECM) content. We show that CD36, a transmembrane receptor that coordinately modulates multiple protumorigenic phenotypes, including adipocyte differentiation, angiogenesis, cell–ECM interactions, and immune signaling, is greatly repressed in multiple cell types of disease-free stroma associated with high mammographic density and tumor stroma. Using both in vitro and in vivo assays, we show that CD36 repression is necessary and sufficient to recapitulate the above-mentioned phenotypes observed in high mammographic density and tumor tissues. Consistent with a functional role for this coordinated program in tumorigenesis, we observe that clinical outcomes are strongly associated with CD36 expression.

Significance: CD36 simultaneously controls adipocyte content and matrix accumulation and is coordinately repressed in multiple cell types within tumor and high mammographic density stroma, suggesting that activation of this stromal program is an early event in tumorigenesis. Levels of CD36 and extent of mammographic density are both modifiable factors that provide potential for intervention. Cancer Discov; 2(9); 826–39. ©2012 AACR.

Read the Commentary on this article by DeClerck, p. 772.

This article is highlighted in the In This Issue feature, p. 753.

Many tissues are comprised of an epithelium associated with a stroma that is composed of extracellular matrix (ECM) proteins, adipocytes, fibroblasts, endothelial cells, neuronal cells, and immune cells. Fibroblasts, a major stromal cell type, produce many components of the ECM, as well as proteases that remodel it (1). In addition, fibroblast signaling is a critical determinant of epithelial and stromal cell fate during development and differentiation (2), tissue homeostasis, and wound healing (1).

Stromal changes associated with malignant lesions are heterogeneous and range from tumor suppressing to tumor promoting. Tumors, subtyped by marker expression in epithelial cells, can be further categorized by their stromal signature, which dictates good or poor outcome (3). Fibroblasts within the stroma of malignant lesions, called carcinoma-associated fibroblasts (CAF), differ from their counterparts in disease-free tissue (4). CAFs can stimulate tumor progression of initiated nontumorigenic epithelial cells, both in vitro andin vivo, whereas normal fibroblasts cannot (5). CAFs increase proliferation and decrease apoptosis in adjacent epithelial cells and promote angiogenesis by recruiting endothelial progenitor cells into tumors (5, 6). All other stromal components, including immune cells and endothelial cells, also participate in malignant progression (4).

The stroma within and immediately adjacent to a malignant lesion may exhibit a range of histologic alterations, called desmoplasia. The alterations range from a predominantly cellular stroma, containing fibroblasts, vascular cells, and immune cells with little ECM, to a dense tissue with a minimum of cells and a maximum content of matrix (4, 7). In addition, tumor stroma exhibits a reduction in the size and number of adipocytes (8). The most odious stromal changes include extreme ECM deposition and remodeling, along with aberrant vasculature and fibroblast and immune cell infiltration. Classically, participation of the stroma has been viewed as a reactive process, in which signals from malignant epithelial cells recruit and stimulate these stromal components.

Interestingly, some desmoplastic features are seen in cancer-free tissues of women at high risk for breast cancer, namely in the context of wound healing, radiation response, pregnancy-associated involution, and high mammographic density (4, 8–10). Mammographic density is of particular interest, as almost one third of breast cancers are thought to be attributable to phenotypes associated with high breast density, and mammographic density is a strong risk factor with high prevalence (11, 12). Mammographic density is determined by the relative amounts of radiolucent material (fat) and radiodense material (epithelial cells, fibroblasts, and connective tissue) within the breast on a mammogram, either of which may occupy anywhere between 0% and 100% of the gland. Radiodense areas exhibit several histologic characteristics of stroma associated with malignant epithelial cells, specifically low adipocyte content and high ECM and stromal cell content (9, 10). Epidemiologic studies suggest that women with high mammographic density have a 4- to 6-fold increase in risk for invasive breast cancer compared with women with low mammographic density (11, 13–15).

Studies comparing pairs of monozygotic and dizygotic twins attributed 53% to 67% of the variation in mammographic density to heritable factors, with the remaining portion modulated by environmental or physiologic factors (16–21). For example, postmenopausal hormone therapy increases mammographic density in some women, whereas tamoxifen treatment decreases mammographic density in both pre- and postmenopausal women and subsequently modulates breast cancer risk (19, 21). This modulation by exogenous factors provides an exciting opportunity for intervention and prevention.

Boyd and colleagues examined the molecular composition of low and high mammographic density tissues (9, 10) and reported higher levels of insulin-like growth factor-I and tissue inhibitor of metalloproteinase 3 in high mammographic density tissue. Recent studies found additional changes including differences in estrogen and prostaglandin metabolism and TGF-β signaling (22–24). Although these studies begin to define the molecular constituents of mammographic density that may explain the basis of high mammographic density and its association with increased breast cancer risk, they are correlative and thus cannot show that these genes or proteins mechanistically modulate phenotypes of mammographic density.

The observation of multiple desmoplastic phenotypes in tissues of high mammographic density, in the absence of a tumor, strongly suggests that these stromal phenotypes are not simply part of a reactive response to existing tumor cells but also represent a program involving multiple stromal components that may create a proactive milieu for the emergence or progression of cancer. Therefore, we hypothesized that the phenotypes observed in both high mammographic density and tumor stroma may be controlled by a common molecular program and thus their stromal components would share phenotypic, molecular, and functional characteristics.

Here we show that the level of expression of a single molecule, CD36, is necessary and sufficient to simultaneously control adipocyte content and matrix accumulation, 2 phenotypes shared by desmoplasia and high mammographic density. CD36, a widely expressed transmembrane receptor, modulates cell type- and ligand-specific phenotypes, including adipocyte differentiation, angiogenesis, apoptosis, TGF-β activation, cell–ECM interactions, and immune signaling (25). We found that CD36 expression was negligible in multiple cell types of tumor stroma compared with surrounding histologically disease-free tissue. Strikingly, high mammographic density tissues, devoid of any malignancy, also showed reduced CD36 levels in multiple cell types compared with low mammographic density tissues. The low level of CD36 expression shared by high-risk (but cancer-free) and malignant tissues suggests that it constitutes a causal and very early event in generating the distinctive characteristics of tumor stroma.

Desmoplasia and Tissues with High Mammographic Density Share Histologic Characteristics

The appearance of stroma associated with malignant epithelial cells is strikingly different from the appearance of most stroma associated with disease-free ductal tissue. Increases in ECM, endothelial and immune components, and the absence of adipocytes are obvious upon inspection (Fig. 1). Strikingly, similar stromal alterations are observed histologically in disease-free tissue of women with high mammographic density as compared with women with low mammographic density, specifically low adipocyte and high ECM content (Fig. 1). We examined fibroblasts from tissue of high and low mammographic density, as well as from tumor tissue, for phenotypic similarities that may underlie their shared histologic appearance.

HDAFs and LDAFs Recapitulate Phenotypes of High and Low Mammographic Density Tissues, Respectively, In Vitro

To determine whether fibroblasts obtained from low and high mammographic density tissues and propagated in vitro could recapitulate the 2 prominent in vivo phenotypic differences described above, low-density associated fibroblasts (LDAF) and high-density associated fibroblasts (HDAF) were purified from disease-free breast biopsies from women with 25% to 50% mammographic density and women with more than 75% mammographic density, respectively (Supplementary Table S1). LDAFs and HDAFs were placed under proliferative or adipocyte differentiation conditions for 6 days before assessing their ability to accumulate fat by Oil Red O staining (Fig. 2A). Fat accumulation was quantitated on a cell-by-cell basis as previously described (Supplementary Fig. S1; ref. 26). This analysis showed that, although both LDAFs and HDAFs accumulated fat under differentiation conditions (10.0-fold and 5.6-fold, respectively, P < 10−8), HDAFs accumulated significantly less fat (3.1-fold, P < 10−8) than LDAFs (Fig. 2B). This decrease in the average amount of fat accumulation per cell was due to a decrease in both the percentage of cells that contain fat and the amount of fat per cell (Supplementary Fig. S2).

To determine whether LDAFs and HDAFs differed in their adipocyte differentiation capacity, rather than simply fat storage, the expression levels of CD36 and leptin (LEP; previously reported to be upregulated during adipocyte differentiation; ref. 27) were measured by quantitative PCR (QPCR) in cells grown under proliferative or adipocyte differentiation conditions for 2 weeks. Both LDAFs and HDAFs could undergo adipocyte differentiation; however, LDAFs showed higher levels of expression of CD36 and LEP compared with HDAFs (CD36: 2.7-fold, P = 0.03; LEP: 6.8-fold, P = 0.05; Fig. 2C). These data showed that differences in fat accumulation between LDAFs and HDAFs are due to differences in their adipocyte differentiation capabilities.

Having shown that cultured LDAFs and HDAFs recapitulated their respective in vivo adipocyte phenotypes, we examined whether they could also recapitulate their in vivo matrix accumulation phenotypes in a short-term assay. LDAF and HDAF cultures were probed after 5 days of propagation for accumulation of matrix proteins known to be increased in desmoplasia: collagen 1A1 (COL1A1), fibronectin (FN1), osteopontin (OPN), and tenascin C (TNC; ref. 1; Supplementary Fig. S3). Cell-by-cell quantitative immunofluorescence analysis (Supplementary Fig. S1) revealed heterogeneity within the cultures with an overall small but statistically significant greater amount of COL1A1, FN1, and OPN (each 1.1-fold, P < 10−8) in HDAFs compared with LDAFs (Fig. 2D). However, HDAFs accumulated a statistically significant lesser amount of TNC (1.1-fold, P < 10−8) than LDAFs, suggesting that there is not a global increase in all matrix protein accumulation (Fig. 2D). One could speculate that this modest increase in matrix accumulation over a prolonged period of time (years), in contrast to our 5-day assay, could account for the dramatic differences observed in vivo.

CAFs Recapitulate Phenotypes of Desmoplasia In Vitro

CAFs have been shown to promote ECM deposition in vitro and in vivo, but little is known about their ability to modulate adipocyte differentiation (1). To this end, human mammary fibroblasts purified from invasive cancer tissues (CAFs) and from reduction mammoplasty tissues (RMF; Supplementary Table S2) were assayed as above for fat accumulation (Fig. 2E, left panel). Cell-by-cell quantitation of Oil Red O staining (Supplementary Fig. S1) showed that, although both RMFs and CAFs accumulated fat under differentiation conditions (27-fold and 8-fold, respectively, P < 10−8), CAFs were deficient in fat accumulation compared with RMFs (5.8-fold, P < 10−8; Fig. 2E, right panel).

CD36 Expression Is Decreased in Fibroblasts from High Mammographic Density and Tumor Tissues

To gain insight into the molecular basis for these phenotypes, gene expression profiling was used to identify genes differentially expressed between LDAFs and HDAFs (GEO GSE38506). CD36 showed the largest difference, its expression being downregulated 3.7- to 4.3-fold in HDAFs relative to LDAFs (Supplementary Table S3). To validate these microarray results, CD36 transcript and protein levels were measured in LDAFs and HDAFs by QPCR and Western blot analysis. Both CD36 gene expression (8.6-fold, P = 0.002) and protein levels were decreased in HDAFs relative to LDAFs (Fig. 3A and B).

Given the striking downregulation of CD36 in fibroblasts of high mammographic density, we examined CD36 expression in fibroblasts isolated from malignant tissue and in microarrays of tumor tissue. CD36 expression, measured by QPCR, was decreased 4.3-fold (P = 0.001) in CAFs relative to RMFs (Fig. 3C). Similarly, analysis of 6 independent public gene expression datasets (Supplementary Table S4) consistently showed a statistically significant decrease in CD36 expression in invasive ductal carcinoma (IDC) tissues compared with normal tissues (3.6- to 12.3-fold). In light of the striking repression of CD36 in fibroblasts from high mammographic density and tumor tissues, we examined the role of CD36 in modulating the 2 prominent in vivo phenotypes shared by these tissues: decreased adipocyte content and increased matrix accumulation (4, 8–10).

CD36 Expression Is Necessary and Sufficient to Modulate Adipocyte Differentiation and Matrix Accumulation In Vitro

Little is known about the basis of decreased adipocyte content in high mammographic density and desmoplastic tissues. However, induction of CD36 expression is necessary for proper adipocyte differentiation in preadipocytes (28) and serves as a marker for terminal differentiation (27). To determine whether CD36 expression could modulate adipocyte content, we genetically modulated CD36 levels in RMFs selected for their intermediate baseline expression level of CD36. To determine whether CD36 expression was necessary for fat accumulation, CD36 expression was repressed in RMFs by transduction of lentiviral particles expressing short hairpin RNA to CD36 (shCD36; Supplementary Fig. S4). Cell-by-cell quantitation of Oil Red O staining (Supplementary Fig. S1) showed that although both control (shLuc) and shCD36 cells accumulated fat under differentiation conditions (8.9-fold and 3.1-fold, respectively, P < 10−8), control cells accumulated significantly more fat than shCD36 cells (6.1-fold, P < 10−8; Fig. 4A). These data showed that CD36 is necessary for fat accumulation. To determine whether CD36 expression was sufficient for fat accumulation, CD36 expression was increased in RMFs by transduction of lentiviral particles expressing CD36 (CD36 OE; Supplementary Fig. S4). Cell-by-cell quantitation of Oil Red O staining (Supplementary Fig. S1) showed that, although both control (vector) and CD36 OE cells accumulated fat under differentiation conditions (2.2-fold and 4.7-fold, respectively, P < 10−8), CD36 OE cells accumulated significantly more fat than control cells (3.3-fold, P < 10−8; Fig. 4B). These data showed that CD36 is sufficient for fat accumulation. Therefore, the decrease in CD36 expression in HDAFs and CAFs could account for their reduced ability to accumulate fat compared with LDAFs and RMFs, respectively.

To determine whether CD36 expression could also modulate matrix accumulation, the fibroblasts with genetically modulated levels of CD36 described above were probed for the accumulation of matrix proteins known to be increased in HDAFs: COL1A1, FN1, and OPN (Fig. 4C). Cell-by-cell quantitative immunofluorescence analysis(Supplementary Fig. S1) revealed that cells with low CD36 (shCD36) accumulated a statistically significant greater amount of COL1A1, FN1, and OPN (1.6-fold, 2.8-fold, and 1.7-fold, respectively, P < 10−8) compared with control cells (shLuc; Fig. 4D, left panel). Conversely, cells with high CD36 (CD36 OE) accumulated a statistically significant lesser amount of COL1A1, FN1, and OPN (1.4-fold, 2.1-fold, and 1.3-fold, respectively, P < 10−8) compared with control cells (vector; Fig. 4D, right panel). Of particular note, these changes in matrix accumulation were already observed after only 5 days of culture. These data showed that reduced CD36 expression is necessary and sufficient for increased matrix accumulation. Taken together, the decreased CD36 expression seen in high mammographic density and tumor tissues can result in the coordinated decrease in fat accumulation and increase in matrix accumulation observed in vivo.

Cd36 Knockout Mice Phenocopy Human High Mammographic Density and Desmoplastic Tissues

To determine whether the modulation of fat and matrix accumulation by CD36, which we observed in vitro, also occurs in vivo, we compared characteristics of fat and matrix accumulation in mammary glands of wild type (WT) and Cd36 knockout (CD36 KO) mice (29). Paraffin sections of the #4 mammary glands of WT or CD36 KO mice were stained with Masson Trichrome, imaged (Fig. 5A) and quantitated on a cell-by-cell (for fat analysis) or image-by-image (for matrix analysis) basis (Supplementary Fig. S1). CD36 KO mice exhibited a statistically significant decrease (1.2-fold, P < 10−8) in fat accumulation (fat cell area; Fig. 5B) and a statistically significant increase (2.1-fold, P < 10−8) in matrix accumulation (Fig. 5C) relative to WT mice. Paraffin sections were stained for COL1A1 and FN1 (Fig. 5D and E, left panels), 2 matrix proteins modulated by CD36 in vitro (Fig. 4C and D). Image-by-image quantitative analysis (Supplementary Fig. S1) revealed that CD36 KO mice accumulated a statistically significant greater amount of COL1A1 and FN1 (both 1.6-fold, P < 10−8 and P = 5.5 × 10−7, respectively) compared with WT mice (Fig. 5D and E, right panels). Therefore CD36 KO mice exhibited 2 prominent phenotypes of high mammographic density and desmoplastic tissues in women: decreased fat accumulation and increased matrix accumulation, showing causality between CD36 expression and modulation of both phenotypes in vitro and in vivo.

CD36 Expression Is Decreased in Multiple Cellular Compartments of High Mammographic Density Tissues Compared with Low Mammographic Density Tissues

CD36 is expressed on the surface of many cells that reside within tissue stroma, but not on epithelial cells (25). Given that microarray analysis of whole tumor tissues showed a dramatic reduction in CD36 compared with disease-free tissues (Supplementary Table S4), we determined whether cellular components other than fibroblasts also exhibited a repression of CD36. Tissue sections from core or excisional biopsies, obtained from cancer-free women with either low mammographic density or with high mammographic density (Supplementary Table S5) were stained for CD36 (Supplementary Fig. S5) and imaged (Fig. 6A, left panels). As expected, adipocytes, endothelial cells, macrophages, and fibroblasts each stained positive for CD36, whereas epithelial cells did not. However, a striking decrease in the intensity of CD36 staining in the high mammographic density tissues was observed in every cell type noted above, indicating that CD36 repression reflects a widespread multicellular program. CD36-positive staining was measured on a cell-by-cell basis as previously described (ref. 26; Supplementary Fig. S1). The average amount of CD36 staining per cell was lower in high mammographic density tissues compared with low mammographic density tissues (4.5-fold, P < 10−8; Fig. 6A, right panel) due to a decrease in both the percentage of CD36-positive cells and the intensity of CD36 staining per cell (Supplementary Fig. S6).

CD36 Expression Is Decreased in Multiple Cell Types within the Stroma of Malignant Lesions Compared with Histologically Normal Adjacent Tissues

Having observed that CD36 expression is decreased in multiple cell types within tissues of high mammographic density, we investigated whether the same would be true in tumor tissues obtained from women with estrogen receptor (ER)/progesterone receptor (PR)-positive, HER2-positive, or ER/PR/HER2-negative (triple negative) IDC (Supplementary Table S6). Tissue sections from this cohort were stained for CD36 and imaged in both the tumor field and the histologically normal adjacent areas (Fig. 6B). In all tumor subtypes, CD36 expression was strikingly absent in the tumor field compared with the normal adjacent tissue (Fig. 6B). Cell-by-cell quantitation of CD36 staining (Supplementary Fig. S1) revealed that CD36 expression was reduced 14.7-fold (P < 10−8) in tumor tissue compared with normal adjacent tissue (Supplementary Fig. S7). This was due to a decrease in both the percentage of CD36-positive cells and the intensity of CD36 staining per cell (Supplementary Fig. S6). A similar dramatic decrease in CD36 expression was observed in each tumor subtype (Fig. 6C). Notably, it has been reported that women with high mammographic density have a higher propensity to develop ER-negative tumors compared with women with low mammographic density (30). Strikingly, normal adjacent tissue from triple negative (ER-negative) tumors exhibited significantly lower levels of CD36 expression compared with ER-positive or HER2-positive tumors (Fig. 6C). Taken together, these data suggested that low expression levels of CD36, as seen in tissues with high mammographic density, might predispose a subset of women to develop triple negative tumors, a tumor subtype associated with poor disease outcome.

To identify the cellular basis of the reduced CD36 staining, we examined the cells that typically express CD36 in histologically normal tissues (adipocytes, endothelial cells, and macrophages). Although scarce, adipocytes within tumor tissue not only showed decreased expression of CD36 but were also smaller in size compared with adipocytes adjacent to the tumor (Supplementary Fig. S7). Staining of serial sections for endothelial and macrophage markers CD31 and CD68, respectively, revealed that endothelial cells and macrophages were still present within the field of tumor tissue, albeit with repressed CD36 expression (Supplementary Fig. S7). In contrast, in the normal adjacent tissue, endothelial cells and macrophages positive for CD31 and CD68, respectively, also stained positive for CD36 (Supplementary Fig. S7). Together with the results in Fig. 3C, these data showed that CD36 expression is actively repressed in fibroblasts, adipocytes, endothelial cells, and macrophages within tumor tissue compared with normal adjacent tissue. This documents that CD36 repression in tumor tissue, also seen in high mammographic density tissue, reflects a widespread, coordinated, and multicellular program.

The repression of CD36 in multiple cell types in high mammographic density tissue compared with low mammographic density tissue, in the absence of a tumor, suggests that this repression can be an early event that precedes overt tumor formation. If this were the case, one might predict that CD36 levels would be repressed in histologically normal tissue directly adjacent to the tumor compared with histologically normal tissue distal to the tumor. To test this hypothesis, multiple tissue sections containing both tumor and histologically normal adjacent tissue or histologically normal distal tissue (>40 mm away from the tumor) were obtained from 3 women with invasive cancer (Supplementary Table S7 and Fig. S8). Tissue sections were stained for CD36, imaged in the tumor, normal adjacent, and normal distal fields (Fig. 7A–C, left panels) and quantified for CD36 expression on a cell-by-cell basis (Supplementary Fig. S1). In accordance with the results reported above, CD36 expression was reduced in the tumor compared with normal adjacent and normal distal tissues in all 3 women (4.7-fold to 113-fold and 6.8-fold to 159-fold, respectively, P < 10−8; Fig. 7A–C, right panels). Consistent with the hypothesis above, we observed in 2 of 3 cases that the normal adjacent tissue exhibited CD36 repression compared with its matched normal distal tissue (1.4-fold, P < 10−8; Fig. 7A and B). In further support of this interpretation, it has been reported that invasive cancers often develop in regions previously measured as radiodense for many years before tumor detection (31). Interestingly, one sample did not show this difference between normal adjacent and normal distal tissues (Fig. 7C). A more detailed analysis of this heterogeneity is currently under investigation.

Clinical Outcome Is Associated with CD36 Expression

Given the dramatic universal repression of CD36 observed within the stroma of all tumor subtypes examined (Fig. 6B and C), we determined whether the initial CD36 levels in nondiseased breast tissue (i.e., mammographic density) would stratify tumors for their ultimate CD36 repression and subsequent clinical phenotypes. Tissue sections containing tumor tissue obtained from women with low mammographic density and women with high mammographic density diagnosed with ER-positive IDC (Supplementary Table S8) were stained for CD36 and imaged (Fig. 6D, left panels). As expected, CD36 staining was very low in all tumor tissues. However, using cell-by-cell quantitation (Supplementary Fig. S1), the average amount of CD36 staining per cell was lower in the tumor tissues of women with high mammographic density compared with women with low mammographic density (1.7-fold, P < 10−8; Fig. 6D, right panel). This was due to a decrease in both the percentage of CD36-positive cells and the intensity of CD36 staining per cell (Supplementary Fig. S6).

Women with high mammographic density who develop invasive cancer have an increased risk for aggressive disease, in that they develop larger tumors of more advanced stage and higher grade compared with women with low mammographic density (30, 32, 33). Because CD36 expression is reduced in tissues of high mammographic density, we examined whether CD36 expression was correlated with tumor grade and size. We interrogated CD36 expression in 2 public gene expression datasets (GSE6532 and GSE9195) for which molecular and associated clinical data were available for tumor samples from a total of 398 treatment-naive IDC patients (34, 35). We analyzed tumor grade, tumor size, and age as a function of CD36 expression using linear regression analyses (Supplementary Table S9). For the purpose of this analysis, we transformed CD36 levels so that they approximated a standard normal distribution (mean ∼0, SD ∼1). These analyses revealed a statistically significant inverse relationship between CD36 expression and tumor grade and, independently, tumor size. CD36 expression was reduced 0.38 SD (P = 0.007) in poorly differentiated tumors (grade 3) compared with well-differentiated tumors (grade 1), after adjusting for the association between CD36 expression and tumor size. In addition, CD36 expression was reduced 0.15 SD (P = 0.003) for every 1 cm increase in tumor size, after adjusting for the association between CD36 expression and tumor grade. There was no significant association between CD36 expression and age.

In sum, stromal characteristics before tumor formation, in this case reflected by extent of mammographic density and CD36 expression levels, are associated with clinical outcome.

Using a combination of in vitro cell assays and in vivo screening of human and murine mammary tissue samples, we show that the level of expression of a single molecule, CD36, is necessary and sufficient to simultaneously control adipocyte content and matrix accumulation, 2 phenotypes that histologically define mammographic density. This is the first report of a gene shown to mechanistically modulate selected phenotypes of mammographic density. An important and novel finding of our study is that repression of CD36 is observed in both stroma associated with a malignant lesion and in breast tissue with high mammographic density in the absence of malignancy. Equally important is our demonstration that repression of CD36 is not limited to a specific cell type but rather is observed in all stromal components, as documented here for adipocytes, endothelial cells, macrophages, and fibroblasts, thus highlighting that expression of CD36 is part of a complex, coordinated, multicellular program. Taken together, these data suggest that expression of this stromal program is not only a reactive response to existing tumor cells but also represents a physiologic state that increases cancer risk. Our data suggest a functional role for this coordinated program because the more repressed CD36 expression is, the more aggressive the tumor.

CD36 is uniquely poised to coordinately orchestrate distinct protumorigenic phenotypes involving multiple cell types. In endothelial cells, in which CD36 blocks VEGF-induced proliferation, migration, and sprouting as well as induces apoptosis in response to ligand binding, its reduced expression could promote a proangiogenic phenotype (36, 37). In activated macrophages, in which CD36 promotes the formation of foam cells and atherosclerotic plaques, triggers the release of inflammatory cytokines and reactive oxygen species (ROS), and inhibits migration (25), its reduced expression could induce multiple, potentially protumorigenic, phenotypes. In support of this, macrophages from CD36 KO mice have reduced levels of ROS, iNOS, and inflammatory cytokines (including IFN-γ and TNF-α) and increased arginase activity (38), indicative of a shift from an M1 proinflammatory/antitumorigenic state to an M2 anti-inflammatory/protumorigenic state (39). In dendritic cells, in which CD36 normally mediates the uptake of apoptotic cells and the cross-presentation of tumor-specific antigens to cytotoxic T cells (25), its reduced expression could allow the tumor to evade the immune system (40). The decreased adipocyte differentiation [with reduced adiponectin (Apn)] and increased matrix accumulation [with increased collagen (Col1a1)] that we observed with reduced CD36 expression in vitro and in vivo have been shown to cause an increase in tumor burden, angiogenesis, and metastasis in a murine mammary tumor model (MMTV-PyMT mice crossed with Apn KO or Col1a1tmJae mice, respectively; refs. 41, 42). Finally, increased matrix accumulation could stiffen the microenvironment, thereby potentially altering mechanosensory networks involved in differentiation and malignancy (43, 44). In light of the central role of CD36 in modulating many functions in multiple cell types, one would predict that the reduction of CD36 expression could have widespread consequences.

The pro-oncogenic tissue state created by repression of CD36 goes far beyond merely supporting the malignant program of a transformed epithelial cell; it creates an interactive milieu that actively participates in tumorigenesis. Hence, our study provides novel insights into a molecular program that controls several stromal phenotypes underlying tumorigenesis. The “single hit, multiple target” nature of CD36, similar to miRNAs, makes it a very attractive therapeutic target. Moreover, in light of its role in mammographic density, strategies to modulate CD36 have the potential to prevent cancer progression in women who are at high risk. In support of this, we show that increased expression of CD36 can restore stromal phenotypes associated with low-risk tissues (Fig. 4). Promisingly, expression of CD36 can be increased in vitro by treatment with aspirin, dexamethasone, statins, or adalimumab and in vivo by treatment with tamoxifen (45–49). Finally, mammographic density can also be modulated by a variety of exogenous agents such as hormones, tamoxifen, and diet. As such, mammographic density represents a highly promising target for intervention as well as a biomarker that can assess response to prevention interventions (19, 21).

Isolation and Propagation of Human Mammary Fibroblasts

Fibroblasts were isolated from reduction mammoplasty, low and high mammographic density, or tumor tissues and assessed for purity as described (5).

Microarray and Quantitative PCR Analysis

Total RNA, isolated from fibroblasts subjected or not to adipocyte differentiation, was used to generate cDNA using TaqMan Reverse Transcription Reagents (Applied Biosystems). The cDNA was used for QPCR (TaqMan) using the standard curve method with primer probe sets for CD36 and leptin. Expression of beta-D-glucuronidase (GUSB) was used to normalize input cDNA. For microarray experiments, mRNA, amplified and biotinylated from total RNA obtained from LDAFs and HDAFs using the MessageAmp II Kit (Ambion), was fragmented and hybridized to HU133 plus 2 chips (Affymetrix).

Adipocyte Differentiation Assay

Fibroblasts, plated at near confluency and grown for 48 hours, were treated with 10 μmol/L 15-deoxy-Δ12,14-Prostaglandin J2 (PJ2; Cayman Chemical) to induce adipocyte differentiation. Lipid accumulation was assayed by Oil Red O staining. Adipocyte differentiation was assessed by monitoring CD36 and leptin expression after 2 weeks of treatment.

Matrix Accumulation

Fibroblasts, plated at near confluency into 8-well glass chamber slides, were grown for 5 days. Matrix accumulation was assessed by monitoring expression of COL1A1 (Abcam), FN1 (BD Biosciences), OPN (Abcam), and TNC (R&D) with primary antibodies all diluted 1:100.

Human and Mouse Tissue Preparation and Immunohistochemistry

Mouse sections (5 μm) were stained with Masson Trichrome using standard protocols. Human sections (4 μm) and mouse sections were subjected to antigen retrieval: microwaving in EDTA buffer (pH = 8.0) for 1 minute for CD36, incubation in Citrate buffer (pH = 6.0) at 80°C for 1 hour for CD31 and CD68, or microwaving in citrate buffer (pH = 6.0) for 10 minutes for COL1A1 and FN1. Sections were incubated for 1 hour with antibodies to CD36 (Sigma-Aldrich), CD31 (Dako), CD68 (Dako), COL1A1 (Abcam), or FN1 (Epitomics) diluted 1:100, 1:20, 1:200, 1:500, and 1:250, respectively. Human sections were incubated for 20 minutes with Primary Antibody Enhancer (Thermo Scientific). All sections were incubated for 30 minutes with HRP Polymer (Thermo Scientific) and 5 minutes with diaminobenzidine, and then counterstained in Mayer hematoxylin to visualize nuclei.

Analysis of Published Microarray Data

Microarray datasets GSE6532 and GSE9195 were analyzed using the robust multiarray average method (50). Expression values for each of the 5 CD36 probe sets were centered by the median, then divided by the SD. Linear regression analyses were used to test the associations between the expression of each CD36 probe set and tumor grade, tumor size, and age. Data from the probe set that were most strongly associated with tumor grade and tumor size are presented in Supplementary Table S9.

Detailed experimental procedures are included in Supplementary Materials and Methods.

No potential conflicts of interest were disclosed.

Conception and design: R.A. DeFilippis, N. Dumont, Y.-Y. Chen, H.K. Berman, K. Kerlikowske, T.D. Tlsty

Development of methodology: R.A. DeFilippis, H.K. Berman, M.L. Gauthier, J. Zhao, T.D. Tlsty

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.A. DeFilippis, J.T. Rabban, Y.-Y. Chen, M.L. Gauthier, J. Zhao, J.J. Marx, J.A. Tjoe, M. Febbraio, K. Kerlikowske, T.D. Tlsty

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.A. DeFilippis, H. Chang, J.T. Rabban, Y.-Y. Chen, H.K. Berman, D. Hu, E. Ziv, K. Kerlikowske, B. Parvin

Writing, review, and/or revision of the manuscript: R.A. DeFilippis, N. Dumont, J.T. Rabban, Y.-Y. Chen, D. Hu, J.J. Marx, J.A. Tjoe, E. Ziv, M. Febbraio, K. Kerlikowske, B. Parvin, T.D. Tlsty

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.A. DeFilippis, G.V. Fontenay, J. Zhao, J.A. Tjoe, E. Ziv, K. Kerlikowske

Study supervision: R.A. DeFilippis, Y.-Y. Chen, T.D. Tlsty

The authors thank members of the Tlsty laboratory, particularly Drs. Colleen Fordyce and Philippe Gascard, for thoughtful discussions, the Nikon Imaging Center (UCSF) for imaging support, and Dr. Alfred Au (UCSF), the breast surgeons (Aurora Health Care), and Carol Halliday, RN (Aurora Health Care), for providing tissue.

This work was supported by NIH/NCI PO1 CA107584 to T.D. Tlsty, K. Kerlikowske, E. Ziv, and B. Parvin (under LBNL contract No. DE-AC02-05CH11231), California Breast Cancer Research Program grant 14OB-0165 to T.D. Tlsty, and NIH/NCI U54 CA143803 to T.D. Tlsty and B. Parvin (UC Riverside).

1.
Kalluri
R
,
Zeisberg
M
. 
Fibroblasts in cancer
.
Nat Rev Cancer
2006
;
6
:
392
401
.
2.
Arendt
LM
,
Rudnick
JA
,
Keller
PJ
,
Kuperwasser
C
. 
Stroma in breast development and disease
.
Semin Cell Dev Biol
2010
;
21
:
11
8
.
3.
Finak
G
,
Bertos
N
,
Pepin
F
,
Sadekova
S
,
Souleimanova
M
,
Zhao
H
, et al
. 
Stromal gene expression predicts clinical outcome in breast cancer
.
Nat Med
2008
;
14
:
518
27
.
4.
Tlsty
TD
,
Coussens
LM
. 
Tumor stroma and regulation of cancer development
.
Annu Rev Pathol
2006
;
1
:
119
50
.
5.
Olumi
AF
,
Grossfeld
GD
,
Hayward
SW
,
Carroll
PR
,
Tlsty
TD
,
Cunha
GR
. 
Carcinoma-associated fibroblasts direct tumor progression of initiated human prostatic epithelium
.
Cancer Res
1999
;
59
:
5002
11
.
6.
Orimo
A
,
Gupta
PB
,
Sgroi
DC
,
Arenzana-Seisdedos
F
,
Delaunay
T
,
Naeem
R
, et al
. 
Stromal fibroblasts present in invasive human breast carcinomas promote tumor growth and angiogenesis through elevated SDF-1/CXCL12 secretion
.
Cell
2005
;
121
:
335
48
.
7.
Walker
RA
. 
The complexities of breast cancer desmoplasia
.
Breast Cancer Res
2001
;
3
:
143
45
.
8.
Bing
C
,
Trayhurn
P
. 
New insights into adipose tissue atrophy in cancer cachexia
.
Proc Nutr Soc
2009
;
68
:
385
92
.
9.
Guo
YP
,
Martin
LJ
,
Hanna
W
,
Banerjee
D
,
Miller
N
,
Fishell
E
, et al
. 
Growth factors and stromal matrix proteins associated with mammographic densities
.
Cancer Epidemiol Biomarkers Prev
2001
;
10
:
243
48
.
10.
Li
T
,
Sun
L
,
Miller
N
,
Nicklee
T
,
Woo
J
,
Hulse-Smith
L
, et al
. 
The association of measured breast tissue characteristics with mammographic density and other risk factors for breast cancer
.
Cancer Epidemiol Biomarkers Prev
2005
;
14
:
343
9
.
11.
Byrne
C
,
Schairer
C
,
Wolfe
J
,
Parekh
N
,
Salane
M
,
Brinton
LA
, et al
. 
Mammographic features and breast cancer risk: effects with time, age, and menopause status
.
J Natl Cancer Inst
1995
;
87
:
1622
9
.
12.
Kerlikowske
K
,
Shepherd
J
,
Creasman
J
,
Tice
JA
,
Ziv
E
,
Cummings
SR
. 
Are breast density and bone mineral density independent risk factors for breast cancer
?
J Natl Cancer Inst
2005
;
97
:
368
74
.
13.
Boyd
NF
,
Byng
JW
,
Jong
RA
,
Fishell
EK
,
Little
LE
,
Miller
AB
, et al
. 
Quantitative classification of mammographic densities and breast cancer risk: results from the Canadian National Breast Screening Study
.
J Natl Cancer Inst
1995
;
87
:
670
5
.
14.
Cummings
SR
,
Tice
JA
,
Bauer
S
,
Browner
WS
,
Cuzick
J
,
Ziv
E
, et al
. 
Prevention of breast cancer in postmenopausal women: approaches to estimating and reducing risk
.
J Natl Cancer Inst
2009
;
101
:
384
98
.
15.
Ziv
E
,
Tice
J
,
Smith-Bindman
R
,
Shepherd
J
,
Cummings
S
,
Kerlikowske
K
. 
Mammographic density and estrogen receptor status of breast cancer
.
Cancer Epidemiol Biomarkers Prev
2004
;
13
:
2090
5
.
16.
Boyd
NF
,
Dite
GS
,
Stone
J
,
Gunasekara
A
,
English
DR
,
McCredie
MR
, et al
. 
Heritability of mammographic density, a risk factor for breast cancer
.
N Engl J Med
2002
;
347
:
886
94
.
17.
Ursin
G
,
Lillie
EO
,
Lee
E
,
Cockburn
M
,
Schork
NJ
,
Cozen
W
, et al
. 
The relative importance of genetics and environment on mammographic density
.
Cancer Epidemiol Biomarkers Prev
2009
;
18
:
102
12
.
18.
Chlebowski
RT
,
McTiernan
A
. 
Biological significance of interventions that change breast density
.
J Natl Cancer Inst
2003
;
95
:
4
5
.
19.
Cuzick
J
,
Warwick
J
,
Pinney
E
,
Duffy
SW
,
Cawthorn
S
,
Howell
A
, et al
. 
Tamoxifen-induced reduction in mammographic density and breast cancer risk reduction: a nested case-control study
.
J Natl Cancer Inst
2011
;
103
:
744
52
.
20.
El-Bastawissi
AY
,
White
E
,
Mandelson
MT
,
Taplin
SH
. 
Reproductive and hormonal factors associated with mammographic breast density by age (United States)
.
Cancer Causes Control
2000
;
11
:
955
63
.
21.
Greendale
GA
,
Reboussin
BA
,
Slone
S
,
Wasilauskas
C
,
Pike
MC
,
Ursin
G
. 
Postmenopausal hormone therapy and change in mammographic density
.
J Natl Cancer Inst
2003
;
95
:
30
7
.
22.
Haakensen
VD
,
Biong
M
,
Lingjaerde
OC
,
Holmen
MM
,
Frantzen
JO
,
Chen
Y
, et al
. 
Expression levels of uridine 5′-diphospho-glucuronosyltransferase genes in breast tissue from healthy women are associated with mammographic density
.
Breast Cancer Res
2010
;
12
:
R65
.
23.
Yang
WT
,
Lewis
MT
,
Hess
K
,
Wong
H
,
Tsimelzon
A
,
Karadag
N
, et al
. 
Decreased TGFbeta signaling and increased COX2 expression in high risk women with increased mammographic breast density
.
Breast Cancer Res Treat
2010
;
119
:
305
14
.
24.
Vachon
CM
,
Sasano
H
,
Ghosh
K
,
Brandt
KR
,
Watson
DA
,
Reynolds
C
, et al
. 
Aromatase immunoreactivity is increased in mammographically dense regions of the breast
.
Breast Cancer Res Treat
2011
;
125
:
243
52
.
25.
Silverstein
RL
,
Febbraio
M
. 
CD36, a scavenger receptor involved in immunity, metabolism, angiogenesis, and behavior
.
Sci Signal
2009
;
2
:
re3
.
26.
Chang
H
,
DeFilippis
RA
,
Tlsty
TD
,
Parvin
B
. 
Graphical methods for quantifying macromolecules through bright field imaging
.
Bioinformatics
2009
;
25
:
1070
5
.
27.
Gregoire
FM
,
Smas
CM
,
Sul
HS
. 
Understanding adipocyte differentiation
.
Physiol Rev
1998
;
78
:
783
809
.
28.
Sfeir
Z
,
Ibrahimi
A
,
Amri
E
,
Grimaldi
P
,
Abumrad
N
. 
CD36 antisense expression in 3T3-F442A preadipocytes
.
Mol Cell Biochem
1999
;
192
:
3
8
.
29.
Febbraio
M
,
Abumrad
NA
,
Hajjar
DP
,
Sharma
K
,
Cheng
W
,
Pearce
SF
, et al
. 
A null mutation in murine CD36 reveals an important role in fatty acid and lipoprotein metabolism
.
J Biol Chem
1999
;
274
:
19055
62
.
30.
Yaghjyan
L
,
Colditz
GA
,
Collins
LC
,
Schnitt
SJ
,
Rosner
B
,
Vachon
C
, et al
. 
Mammographic breast density and subsequent risk of breast cancer in postmenopausal women according to tumor characteristics
.
J Natl Cancer Inst
2011
;
103
:
1179
89
.
31.
Pereira
SM
,
McCormack
VA
,
Hipwell
JH
,
Record
C
,
Wilkinson
LS
,
Moss
SM
, et al
. 
Localized fibroglandular tissue as a predictor of future tumor location within the breast
.
Cancer Epidemiol Biomarkers Prev
2011
;
20
:
1718
25
.
32.
Aiello
EJ
,
Buist
DS
,
White
E
,
Porter
PL
. 
Association between mammographic breast density and breast cancer tumor characteristics
.
Cancer Epidemiol Biomarkers Prev
2005
;
14
:
662
8
.
33.
Kerlikowske
K
,
Cook
AJ
,
Buist
DS
,
Cummings
SR
,
Vachon
C
,
Vacek
P
, et al
. 
Breast cancer risk by breast density, menopause, and postmenopausal hormone therapy use
.
J Clin Oncol
2010
;
28
:
3830
7
.
34.
Loi
S
,
Haibe-Kains
B
,
Desmedt
C
,
Wirapati
P
,
Lallemand
F
,
Tutt
AM
, et al
. 
Predicting prognosis using molecular profiling in estrogen receptor-positive breast cancer treated with tamoxifen
.
BMC Genomics
2008
;
9
:
239
.
35.
Loi
S
,
Haibe-Kains
B
,
Desmedt
C
,
Lallemand
F
,
Tutt
AM
,
Gillet
C
, et al
. 
Definition of clinically distinct molecular subtypes in estrogen receptor-positive breast carcinomas through genomic grade
.
J Clin Oncol
2007
;
25
:
1239
46
.
36.
Primo
L
,
Ferrandi
C
,
Roca
C
,
Marchio
S
,
di Blasio
L
,
Alessio
M
, et al
. 
Identification of CD36 molecular features required for its in vitro angiostatic activity
.
FASEB J
2005
;
19
:
1713
5
.
37.
Jimenez
B
,
Volpert
OV
,
Crawford
SE
,
Febbraio
M
,
Silverstein
RL
,
Bouck
N
. 
Signals leading to apoptosis-dependent inhibition of neovascularization by thrombospondin-1
.
Nat Med
2000
;
6
:
41
8
.
38.
Kennedy
DJ
,
Kuchibhotla
S
,
Westfall
KM
,
Silverstein
RL
,
Morton
RE
,
Febbraio
M
. 
A CD36-dependent pathway enhances macrophage and adipose tissue inflammation and impairs insulin signalling
.
Cardiovasc Res
2011
;
89
:
604
13
.
39.
Sica
A
,
Larghi
P
,
Mancino
A
,
Rubino
L
,
Porta
C
,
Totaro
MG
, et al
. 
Macrophage polarization in tumour progression
.
Semin Cancer Biol
2008
;
18
:
349
55
.
40.
McDonnell
AM
,
Robinson
BW
,
Currie
AJ
. 
Tumor antigen cross-presentation and the dendritic cell: where it all begins
?
Clin Dev Immunol
2010
;
2010
:
539519
.
41.
Landskroner-Eiger
S
,
Qian
B
,
Muise
ES
,
Nawrocki
AR
,
Berger
JP
,
Fine
EJ
, et al
. 
Proangiogenic contribution of adiponectin toward mammary tumor growth in vivo
.
Clin Cancer Res
2009
;
15
:
3265
76
.
42.
Provenzano
PP
,
Inman
DR
,
Eliceiri
KW
,
Knittel
JG
,
Yan
L
,
Rueden
CT
, et al
. 
Collagen density promotes mammary tumor initiation and progression
.
BMC Med
2008
;
6
:
11
.
43.
Engler
AJ
,
Sen
S
,
Sweeney
HL
,
Discher
DE
. 
Matrix elasticity directs stem cell lineage specification
.
Cell
2006
;
126
:
677
89
.
44.
Levental
KR
,
Yu
H
,
Kass
L
,
Lakins
JN
,
Egeblad
M
,
Erler
JT
, et al
. 
Matrix crosslinking forces tumor progression by enhancing integrin signaling
.
Cell
2009
;
139
:
891
906
.
45.
Vinals
M
,
Bermudez
I
,
Llaverias
G
,
Alegret
M
,
Sanchez
RM
,
Vazquez-Carrera
M
, et al
. 
Aspirin increases CD36, SR-BI, and ABCA1 expression in human THP-1 macrophages
.
Cardiovasc Res
2005
;
66
:
141
9
.
46.
Matasic
R
,
Dietz
AB
,
Vuk-Pavlovic
S
. 
Dexamethasone inhibits dendritic cell maturation by redirecting differentiation of a subset of cells
.
J Leukoc Biol
1999
;
66
:
909
14
.
47.
Ruiz-Velasco
N
,
Dominguez
A
,
Vega
MA
. 
Statins upregulate CD36 expression in human monocytes, an effect strengthened when combined with PPAR-gamma ligands Putative contribution of Rho GTPases in statin-induced CD36 expression
.
Biochem Pharmacol
2004
;
67
:
303
13
.
48.
Boyer
JF
,
Balard
P
,
Authier
H
,
Faucon
B
,
Bernad
J
,
Mazieres
B
, et al
. 
Tumor necrosis factor alpha and adalimumab differentially regulate CD36 expression in human monocytes
.
Arthritis Res Ther
2007
;
9
:
R22
.
49.
Saibara
T
,
Ogawa
Y
,
Onishi
S
. 
Tamoxifen in early breast cancer
.
Lancet
1998
;
352
:
404
.
50.
Irizarry
RA
,
Hobbs
B
,
Collin
F
,
Beazer-Barclay
YD
,
Antonellis
KJ
,
Scherf
U
, et al
. 
Exploration, normalization, and summaries of high density oligonucleotide array probe level data
.
Biostatistics
2003
;
4
:
249
64
.