Heparanase Accelerates Obesity-Associated Breast Cancer Progression

Obesity [defined as a body mass index (BMI) above or equal to 30 kg/m²] is a widespread public health problem that has been consistently associated with several pathologies, including at least 13 different types of cancer (1). In particular, numerous studies have demonstrated that obese women have a significantly higher risk of bearing breast tumors, largely estrogen receptor (ER)-positive (2, 3), which are resistant to therapies, more likely to recur, and associated with higher death rates (3–5). As the recently estimated global prevalence of overweight/obesity in women is approximately 40% and postmenopausal breast carcinoma has the highest incidence among the three obesity-related cancers in females (i.e., breast, endometrial, and ovarian; ref. 1), elucidation of the precise molecular mechanisms underlying breast tumor—promoting action of obesity is of high importance.

Obesity is widely recognized as a low-grade chronic inflammatory state, and inflammation is one of the most likely contributors to the obesity–cancer link (4–7). Macrophages are key immunocytes mediating the tumor promotion action of obesity (4–7). In the setting of obesity-associated ER-positive breast carcinoma, the procancerous action is exerted both by tumor-associated macrophages (TAM; ref. 8) and by adipose tissue macrophages (atM; refs. 6, 9–12). Indeed, macrophage mobilization/activation was observed in visceral and mammary adipose in animal models of obesity, as well as in obese patients, correlating with both BMI and adipocyte hypertrophy (4, 5, 9, 11, 13). The atM infiltrating “obese” adipose tissue forms histologically visible crown-like structures (CLS), composed of macrophages surrounding dysfunctional adipocytes, a hallmark of white adipose tissue inflammation (3, 9, 10). Moreover, unlike in “lean” adipose tissue, atM in “obese” adipose tissue acquire proinflammatory phenotype, among other factors—due to activation by fatty acids, derived from the lipid-laden “obese” adipocytes (3, 9, 10, 12). As a result of this activation, reportedly mediated via toll-like receptor 4 (TLR4; refs. 5, 14–16), CLS-residing macrophages secrete inflammatory mediators (i.e., TNFα; refs. 6, 9, 10), which induce expression of aromatase in adipose stromal cells (i.e., fibroblasts; refs. 9, 17, 18). Aromatase is the rate-limiting enzyme in estrogen biosynthesis, which catalyzes the conversion of androgens to estrogens. After the cessation of ovarian function, estrogens in postmenopausal women are synthesized exclusively in peripheral tissues, with adipose being a major site of estrogen production. As obesity-associated breast tumors are mostly ER-positive and occur in postmenopausal patients (2, 3, 11), induction of aromatase expression in “obese” adipose tissue and associated increase in estrogen production, owing to the interplay between “obese” adipocytes, atM, and fibroblasts, are important determinants in breast carcinoma–obesity link (3, 6, 9).

Along with the atM infiltrating excessive adipose tissue, TAM also contribute to obesity-associated breast cancer. When polarized toward tumor-promoting phenotype, TAM supply key obesity-associated procancerous cytokines (i.e., IL6, CCL2), growth factors (e.g., VEGF, EGF), and activating tumor-stimulating signaling pathways (e.g., NFκB, STAT3; refs. 4, 7, 19), thus contributing to breast carcinoma progression (4, 5, 8). Components of the obese milieu, most notably free fatty acids (FFA), whose circulating levels are often increased in obesity, are among the candidate agents responsible for adverse TAM activation (5, 14–16). Nevertheless, given the functional plasticity of macrophages and the notorious ability of TAM to exhibit both anti- and protumor activities (8), the mechanisms of TAM phenotypic switch occurring under obese state are not well elucidated, and it is not clear whether circulating levels of FFA (20) alone are sufficient to drive protumorigenic TAM polarization. Even less is known about obesity-associated alterations in the extracellular matrix (ECM) of breast carcinoma and their role in coupling obese state and breast tumor progression.

Here we provide evidence that heparanase enzyme [the only known mammalian endoglycosidase-degrading heparan sulfate (HS) chains in the ECM] mediates effects of excess adiposity on breast cancer progression. Upregulation of heparanase is documented in clinical and experimental breast carcinoma (21–24), correlating with larger tumor size, poor survival, and resistance to therapy (21, 25). The enzymatic substrate of heparanase, HS, is ubiquitously present at the cell surface and ECM, playing essential roles in ECM integrity and regulation of receptor–ligand interactions that involve a variety of bioactive molecules (26, 27). In particular, intact extracellular HS inhibits TLR4 responses and macrophage activation, whereas its enzymatic removal relieves this inhibition (28). Moreover, soluble HS fragments generated by heparanase (29) stimulate TLR4 signaling in vitro (28, 30) and in vivo (31). Hence, heparanase shapes macrophage responses in several pathophysiologic conditions (30, 32–35).

Incorporating the aforementioned data with observations that heparanase deficiency abolishes obesity-accelerated breast tumor progression in vivo, and that the enzyme is preferentially expressed in both clinical and experimental obesity-associated breast tumors (see “Results” section below), we hypothesized that heparanase mediates the tumor-promoting effect of obesity by directing procancerous action of macrophages at the interface of breast carcinoma with excess adiposity and inflammation.

Breast tumor tissue specimens and clinical data from 123 female patients with breast carcinoma were available from the Sharett Oncology Institute, Hadassah Medical Center (Jerusalem, Israel). A summary of clinical and pathologic characteristics of the study population is shown in Supplementary Table S1. The use of these data and formalin-fixed, paraffin-embedded breast carcinoma tissues in research was approved by the Human Subjects Research Ethics Committee of the Hadassah Medical Center (Jerusalem, Israel). Tissue microarray construction was performed as previously (25). Briefly, 5-μm sections stained with hematoxylin and eosin were obtained to confirm the diagnosis and to identify representative areas of the specimen. From these defined areas, three tissue cores with a diameter of 0.6 mm were taken from the different regions of the tumor and arrayed in triplicates on a recipient paraffin block as described previously (25). Sections of 5 μm of the recipient blocks were cut, deparaffinized, and rehydrated. Tissue was then incubated in 3% H₂O₂, denatured by boiling (3 minutes) in a microwave oven in citrate buffer (0.01 mol/L, pH 6.0) and blocked with 10% goat serum in PBS. Sections were incubated with polyclonal rabbit anti-heparanase antibody (733) directed against a synthetic peptide (¹⁵⁸KKFKNSTYRSSSVD¹⁷¹) corresponding to the N-terminus of the 50-kDa subunit of the heparanase enzyme (25, 33). The antibody was diluted 1:100 in 10% goat serum in PBS. Control slides were incubated with 10% goat serum alone. Color was developed as described in ref. 25, slides were visualized with a Zeiss axioscope microscope, and manually read by an expert pathologist (B. Maly). To define tumor as heparanase-positive, a cut-off point of 25% immunostained tumor cells was chosen on the basis of an initial overview of the cases, to improve signal-to-noise ratios. Cutoff was chosen before any attempt at correlating heparanase expression with the obese status of the patients. Immunodetection of ERα was performed as described in ref. 25.

Human ER-positive breast carcinoma cell lines T47D and MCF7 (authenticated by short tandem repeat profiling at the Genomics Center of the Biomedical Core Facility, Technion University, Haifa, Israel), mouse breast carcinoma E0771 (C57BL/6J syngeneic ER-positive cell line), kindly provided by Dr. R. Sharon, Hebrew University Medical School (Jerusalem, Israel), and mouse pancreatic carcinoma Panc02 (C57BL/6J syngeneic cell line) provided by Dr. M. Dauer (University of Munich, Munich, Germany), were grown in RPMI1640 (T47D, E0771, Panc02) or DMEM (MCF7) supplemented with 1-mmol/L glutamine, 50 μg/mL streptomycin, 50 U/mL penicillin, and 10% FCS (Biological Industries) at 37°C and 8% CO₂. All cell lines were tested routinely for Mycoplasma by the PCR assay (Biological Industries). Prior to estrogen treatment, cells were maintained for 4 days in phenol red-free medium supplemented with charcoal-stripped FCS (Biological Industries). Then, medium was changed to serum-free medium and cells were treated with estrogen or vehicle alone (ethanol), as indicated in the Results section.

Ten-week-old, female wt C57BL/6J mice (Harlan Laboratories) and heparanase-knock out (Hpse-KO) mice (36) on C57BL/6J background (n ≥ 7 per experimental group) were fed high-fat diet (HFD; Teklad TD.06414, 60% of total calories from fat), or control diet (CD; Teklad 2018S) for 15 consecutive weeks. By the end of experimental week 12, when both wt and Hpse-KO HFD-fed animals became obese, E0771 cells were injected orthotopically into fourth left mammary fat pads of both HFD-fed (obese) and CD-fed (lean) wt and Hpse-KO mice (5 × 10⁵ cells per injection). Similarly, HFD-fed (obese) and CD-fed (lean) wt and Hpse-KO mice were injected subcutaneously with Panc02 cells (5 × 10⁵ cells per injection). The tumor volume was monitored until experimental week 15. Animals were then sacrificed and tissue samples collected from tumors, collateral (right) mammary gland adipose and visceral fat, and snap-frozen for RNA/protein extraction or processed for histology. All experiments were performed in accordance with the Hebrew University Institutional Animal Care and Use Committee.

The results are presented as the mean ± SD unless otherwise stated. P values of ≤ 0.05 were considered statistically significant. Statistical analysis of in vitro and in vivo data was performed by unpaired Student t test. Pearson χ² test was applied to analyze the relationship between heparanase expression and obese status of patients with breast carcinoma, using SPSS software (SPSS Inc.). All statistical tests were two-sided.

Formalin-fixed, paraffin-embedded human breast carcinoma tissues and clinical data were obtained in accordance with the ethical guidelines of the Declaration of Helsinki. The studies were approved by the Human Subjects Research Ethics Committee of the Hadassah Medical Center (Jerusalem, Israel). The Committee determined that obtaining a written informed consent from each subject was unnecessary for this study.

Animal experiments were approved by Institutional Animal Care Committee of the Hebrew University (Jerusalem, Israel). Additional methods are presented in the Supplementary Methods.

To investigate a role for heparanase in coupling obesity and breast cancer progression, wild-type (wt) and heparanase-null (Hpse-KO) mice on C57BL6 background were utilized in the model of HFD-induced obesity, as described in Materials and Methods. HFD-fed C57BL6 mice represent one of the most reliable and best studied models of diet-induced obesity and related pathologic changes, including inflammation (37, 38), fatty acid accumulation (39), and obesity-accelerated tumor growth (19, 38). Animals of both genotypes (wt and Hpse-KO) became obese following 12 weeks of HFD, as evidenced by their significantly increased body weight and adipocyte hypertrophy in intra-abdominal fat (characteristic feature of diet-induced obesity, Fig. 1 A and B). On experimental week 12, E0771 cells (syngeneic ER-positive breast cancer cell line; ref. 19) were injected orthotopically into the mammary fat pad of both HFD-fed (obese) and CD-fed (lean) wt and heparanase-KO mice. As expected (19, 38), HFD-induced obesity markedly accelerated tumor progression in wt mice: 2.3-fold larger tumors were observed in wt obese versus wt lean mice on experimental week 15, 3 weeks posttumor inoculation (Fig. 1C). Strikingly, heparanase deficiency in mouse host abolished tumor-accelerating effect of obesity, as demonstrated by the lack of statistically significant difference between the volume of tumors growing in lean versus obese Hpse-KO mice (Fig. 1D). Of note, heparanase deficiency had no inhibitory effect on E0771 tumor growth in nonobese conditions, as no statistically significant difference was detected between the volume of tumors growing in lean wt versus lean Hpse-KO mice.

We next compared heparanase expression in E0771 breast tumors growing in obese versus lean mice, applying immunoblotting and IHC. Heparanase expression was barely detected in E0771 tumors derived from the lean host, whereas a marked increase in heparanase protein levels was detected in tumors derived from obese wt host (Fig. 2A and B; Supplementary Fig. S1). In agreement with the increased expression of heparanase, a significant decrease in the content of HS (enzymatic substrate of heparanase) was detected in the tumors of obese versus lean wt mice, whereas no change in HS content was detected in the tumors of obese versus lean heparanase-KO mice (Supplementary Fig. S2A and S2B).

Next, to validate the clinical relevance of these findings, we utilized tissue microarray comprising 123 tumor specimens derived from obese (BMI ≥30) and nonobese (BMI <30) patients with breast carcinoma and examined the correlation between heparanase expression and obese status of the patients. Because the association between obesity and breast cancer is the highest among hormone-dependent breast tumors (2, 3), ER-positive and ER-negative cases were analyzed separately. χ² analysis was then used to assess the relationship between obese status and upregulation of heparanase. In full agreement with the in vivo data (Fig. 2A and B), strong and significant correlation between obese state of the patient and overexpression of heparanase was noted in ER-positive breast carcinoma: almost 2-fold higher proportion of heparanase-overexpressing tumors was detected in obese versus nonobese ER-positive patients with breast carcinoma (63.2% vs. 36.8%, χ² test; P = 0.045; Fig. 2C).

Because association between increased adiposity and breast carcinoma risk/aggressive behavior was repeatedly demonstrated in overweight (BMI ≥25 kg/m²) patients (1–3), heparanase expression in breast carcinoma tumors derived from overweight versus normal weight women (BMI <25 kg/m²) was examined as well. As shown in Supplementary Fig. S3A, significant correlation between overweight condition and heparanase overexpression was noted in ER-positive patients with breast carcinoma. No association between heparanase expression and either overweight (Supplementary Fig. S3B) or obesity (Supplementary Fig. S3C) was detected in ER-negative tumors. Of note, the enzyme expression was not detected in any of the 5 control specimens of nonmalignant breast tissue, included in the array.

Both carcinoma cells per se and host-derived stromal cells can potentially serve as a source of the enzyme in the microenvironment of breast carcinoma and other tumor types (40–43). As shown in Fig. 2A, under nonobese conditions, E0771 carcinoma cells express very low or no detectable levels of heparanase. This notion, taken together with the fact that elevated levels of heparanase were detected in tumors growing in obese wt mice (but not in obese heparanase-KO mice; Fig. 2A, middle), may lead to the assumption that in this experimental setting heparanase is mainly contributed by the host-derived stromal cells, rather than by carcinoma cells. However, immunostaining of the mouse tumor tissues with heparanase antibody clearly demonstrated that E0771 carcinoma cells per se represent the main source of the enzyme in tumors growing in obese wt mice (Fig. 2B). In a similar manner, in patient-derived ER-positive breast carcinoma specimens, overexpression of heparanase was noted in breast carcinoma cells rather than in stromal elements (Fig. 2D).

Estrogen is a potent inducer of the heparanase enzyme expression in the hormone-responsive cell types (23, 25, 44) including ER-positive breast carcinoma cells of both mouse (i.e., E0771, Fig. 3A) and human (i.e., MCF7, T47D; refs. 23, 25) origins. We therefore next queried whether enhanced estrogen signaling (a known consequence of augmented aromatase production by excess adiposity; refs. 3, 4, 9, 11, 13) is responsible for the induction of heparanase in E0771 breast carcinoma cells growing in obese wt mice. Consistent with this mode of action, we detected approximately 6-fold increased expression of progesterone receptor (estrogen-regulated gene commonly used as an indicator of enhanced estrogen signaling) in E0771 breast tumors growing in obese wt mice but not in obese Hpse-KO mice (Fig. 3B). These data support the notion that obesity-associated enhanced estrogen signaling occurs in wt but not in Hpse-KO animals.

As stated above, obese state initiates series of cellular/molecular events leading to increased production of estrogen in adipose tissue (3, 4, 9, 11, 13), including formation of CLS by atM aggregating around dysfunctional adipocytes (9, 10). These atM undergo abnormal activation by fatty acids released by the lipid-loaded “obese” adipocytes. Notably, female patients store greater amounts of fatty acids (represented by palmitate) in their fat depots, as compared with male patients (45). Palmitate and other saturated fatty acids are capable of activating macrophages in “obese” adipose tissue (acting via TLR4; refs. 5, 9, 10, 12, 14–16). Thus, CLS-residing atM are adversely activated and secrete elevated levels of cytokines (9, 10, 12), including TNFα, a key inducer of aromatase expression in adipose stromal cells–the fibroblasts that surround adipocytes (9, 17, 18), eventually leading to extragonadal production of estrogen (3, 4, 9, 13, 18).

We, therefore, examined the correlation between obese state and aromatase expression in visceral adipose tissue of lean and obese mice. As expected (46), no aromatase expression was detected in adipose tissue in the majority of lean wt mice (Fig. 3C) In contrast, aromatase expression was easily detected in the adipose tissue in >80% of obese wt mice (Fig. 3C). Statistical analysis confirmed correlation between obese state and aromatase expression in wt animals (two-sided χ² test; P = 0.003, Fig. 3C). These findings are consistent with the clinically observed correlation between obese status and induction of aromatase in adipose tissue of obese patients with breast carcinoma, as well as in dietary/genetic models of obesity (9, 11, 13, 46, 47).

Remarkably, obese state did not result in aromatase induction in adipose tissue of heparanase-KO mice (Fig. 3C). Given the role of “obese” adipocyte-derived fatty acids in TLR4-dependent atM activation (5, 9, 10, 12, 14–16), and the involvement of heparanase in sustaining macrophage stimulation via TLR4 (32, 33, 35), we hypothesized that total lack of heparanase does not allow for adequate atM activation in adipose tissue of obese Hpse-KO animals. As a result, atM in obese Hpse-KO mice (as opposed to obese wt mice) may not be able to produce sufficient amounts of inflammatory cytokines (i.e., TNFα) required for the induction of aromatase expression in adipose tissue fibroblasts. To validate this hypothesis, we first examined CLS formation in adipose tissue specimens harvested from lean/obese wt and Hpse-KO mice. Essentially, no CLS were detected in the adipose tissue of both wt and Hpse-KO lean mice (in agreement with ref. 7), whereas in the adipose tissue of obese mice (of both genotypes), CLS were easily detectable [histologically and by the presence of F4/80-positive macrophages (Fig. 4A, left and middle)]. Of note, there was no statistically significant difference between the number of CLS observed in obese wt versus obese Hpse-KO adipose tissue (3.75 vs. 2.67 CLS/500 adipocytes, P = 0.504), suggesting that heparanase deficiency does not affect the degree of macrophage infiltration in adipose tissue. Next, we tested the possibility that heparanase affects the activation state of CLS-residing macrophages (i.e., level of TNFα production). To this end, we utilized immunostaining for F4/80 (mouse macrophage-specific marker) and TNFα (a powerful inducer of aromatase expression in obese adipose tissue; refs. 3, 9, 13, 17, 18) in serial sections of adipose tissue specimens derived from wt and Hpse-KO mice. TNFα immunostaining revealed markedly lower levels of the cytokine in Hpse-KO mice (Fig. 4A, right), and quantification studies confirmed a statistically significant difference in staining intensity of TNFα in CLS macrophages of adipose specimens harvested from obese wt mice, as compared with obese Hpse-KO mice (Fig. 4B).

We next turned to validate in vitro the decreased ability of Hpse-KO macrophages to undergo activation in response to obesogenic stimulation (Fig. 4C). For this purpose, we compared expression of TNFα by macrophages derived from wt and Hpse-KO mice, in response to fatty acid, mimicking macrophage activation by fatty acids present in obese adipose tissue (5, 9, 10, 12, 14–16). Palmitate (the dominant saturated fatty acid present in fat depots of female patients; ref. 45) was utilized in our experiments, along with other saturated (i.e., stearic) and unsaturated (oleic, linoleic) fatty acids.

Consistent with our hypothesis, stimulation with palmitate (C16:0) in vitro strongly induced expression of TNFα in wt macrophages but not in heparanase-deficient macrophages (Fig. 4C). Similar results were obtained following stimulation with stearic acid (C18:0, Fig. 4C). Unsurprisingly, unsaturated fatty acids [oleic (C18:1) and linoleic (C18:2), Fig. 4C] failed to induce significant cytokine expression in wt macrophages, consistent with reports that attributed TLR-activating/macrophage-stimulating properties mainly to saturated fatty acids (i.e., palmitate, stearate), whereas unsaturated acids (i.e., oleate) were reported to display only weak macrophage-stimulating ability (14).

Then, to further test our hypothesis that presence of heparanase is a prerequisite for the ability of fatty acid–stimulated macrophages to undergo activation sufficient to induce aromatase in adipose stromal cells, we utilized fibroblasts isolated from breast adipose tissue of healthy women undergoing reduction mammoplasty (as described in Supplementary Methods), known as a useful and reliable model to study regulation of aromatase expression in adipose tissue (9, 17, 18). The fibroblasts were incubated with medium conditioned by wt or heparanase-deficient macrophages, stimulated by palmitate. In agreement with the hypothesized mode of heparanase action in this system, medium conditioned by wt macrophages, but not by Hpse-KO macrophages, induced expression of aromatase in the fibroblasts (Fig. 4D).

These results, showing that heparanase deficiency abolishes the macrophage ability to induce aromatase expression in adipose fibroblasts ex vivo, together with the in vivo findings depicted in Figs. 3C and 4A–C, suggest an explanation for the lack of aromatase induction in heparanase-deficient obese adipose tissue (Fig. 3C) and therefore the absence of enhanced estrogen signaling (Fig. 3B) and lack of heparanase induction in E0771 tumors growing in obese Hpse-KO mice (Fig. 2A and B).

Although adipose tissue-residing macrophages (atM) are certainly important in obesity-associated ER-positive breast cancer, TAM are undoubtedly the main procancerous immunocyte population in general (4, 5, 8) and in obesity-accelerated breast carcinoma progression in particular (5, 7, 19). Thus, turning back to E0771 tumors, we compared the degree of TAM infiltration in tumor tissue specimens derived from lean versus obese mice, applying immunostaining with antibodies directed against F4/80 (Fig. 5A) and Mac2 (Supplementary Fig. S4A and S4B)—specific mouse macrophage markers (48). Augmented infiltration of TAM was observed in obesity-associated E0771 breast carcinoma growing in wt mice (2-fold increase, P < 0.02—Fig. 5A and B, consistent with refs. 7, 19). Importantly, heparanase deficiency abolished increased macrophage recruitment in obesity-associated tumors, as no statistically significant difference in TAM infiltration was detected in tumors growing in obese versus lean Hpse-KO mice (Fig. 5A and B), mirroring the lack of procancerous action of obesity in these mice (Fig. 1D).

Macrophage infiltration in breast tumors correlates with high expression of C-C chemokine ligand 2 (CCL2). Moreover, CCL2, which can be supplied by TAM themselves, acts as a chief inducer of macrophage mobilization in breast carcinoma tissue (7, 49). Furthermore, CCL2-mediated recruitment of macrophages is a key molecular event underlying obesity-associated breast cancer progression (7). Thus, we compared the expression of CCL2 in tumors derived from obese versus lean mice. Although no difference in CCL2 levels was noted between tumors derived from obese versus lean Hpse-KO mice, significant increase in both CCL2 mRNA (>2-fold increase, P = 0.048) and protein levels (Fig. 5C and D) was detected in tumors derived from wt obese, as compared with wt lean mice. Given that CCL2 is a well-characterized HS-binding chemokine, relatively modest increase in CCL2 protein levels detected by immunostaining (Fig. 5C and D) in obese wt mice (vs. >2-fold increase in CCL2 mRNA) may reflect elevated levels of heparanase enzyme and the resulting decrease in HS content in the tumors from obese versus lean wt mice (Fig. 2A and B; Supplementary Fig. S2A and S2B). In addition, colocalization experiments revealed a 2-fold increase in the percentage of CCL2–positive macrophages infiltrating tumors growing in obese versus lean wt (but not Hpse-KO) mice (Supplementary Fig. S4A and S4B). Collectively, these findings suggest that heparanase overexpression in tumors growing in obese wt mice (Fig. 2A and B) directly affects TAM recruitment and activation.

The aforementioned findings, along with the emerging role of heparanase in sustaining macrophage reactivity in several cancer-related and nonrelated disorders (32–35, 50), lead to hypothesis that in the setting of obesity, increased intratumoral levels of heparanase (Fig. 2) may direct protumorous action of TAM stimulated by obesogenic substances present in circulation. Specifically, FFAs are among the candidate agents capable of modulating macrophage responses in the obese individuals (5, 9, 10, 12, 14–16). However, unlike within the “obese” adipose tissue, where CLS-residing atM are exposed to relatively high levels of fatty acids due to necrosis of hypertrophic adipocytes (9), obesity-associated circulating levels of FFA (20) alone might not be sufficient to trigger TAM polarization.

To examine in vitro the hypothesized effect of heparanase on phenotypic switch of macrophages exposed to obesogenic substances, we utilized oleic acid, taking advantage of the fact that it is one of the most abundant fatty acids in circulation (51) and in the HFD used in this study (see Materials and Methods). In addition, oleic acid per se was shown to induce only slight effects on macrophage activation and induction of cytokine expression (i.e., TNFα, IL6; Fig. 4C; ref. 14). We first tested the effect of the enzyme on the ability of oleic acid–stimulated macrophages to produce protumorous cytokines and to promote breast carcinoma cell proliferation in vitro. To recapitulate conditions faced by macrophages infiltrating obesity-associated breast carcinoma tissue [i.e., exposure to increased levels of heparanase (Fig. 2) along with the presence of elevated levels of circulating FFA], primary macrophages were treated with oleic acid in the presence of recombinant active heparanase; macrophages treated with oleic acid in absence of heparanase or in the presence of heat-inactivated enzyme were used as control. As depicted in Fig. 6A, heparanase strongly and in a dose-dependent manner augmented production of IL6 (a key breast carcinoma–promoting cytokine implicated in the obesity-cancer link; refs. 4–6) by oleic acid–stimulated macrophages. The effect of heparanase on IL6 expression/secretion was dose dependent (Fig. 6A; Supplementary Fig. S5A and S5B); IL6 levels reached a plateau and did not continue to increase at the enzyme concentrations above 3 μg/mL (Supplementary Fig. S5A and S5B). Importantly, heparanase effect was dependent on its enzymatic activity because heat-inactivated heparanase did not affect cytokine levels.

IL6 acts in promotion of cancer through activation of several signaling pathways leading to increased tumor cell proliferation, survival, and resistance to therapies (4, 6, 52, 53). In particular, IL6 was shown to enhance tumor cell proliferation in ER-positive breast cancer (52, 53), and higher IL6 levels in patients with breast carcinoma correlate with poorer survival and diminished response to therapies (4). One important mechanism by which IL6 promotes breast carcinoma is activation of STAT3 (4, 52, 53), a critical component of tumor-stimulating signaling in breast carcinoma and other malignancies. Notably, we found that medium conditioned by macrophages stimulated by oleic acid in the presence of heparanase markedly enhanced both STAT3 signaling (Fig. 6B–D) and proliferation (Fig. 6E) of E0771 cells in vitro. Proliferation of human breast carcinoma cell lines T47D and MCF7 was also enhanced by medium conditioned by macrophages stimulated by oleic acid in the presence of heparanase (Supplementary Fig. S6).

In contrast, medium conditioned by macrophages stimulated by oleic acid in the absence of the enzyme had no effect on STAT3 signaling (Fig. 6B–D) and cell proliferation (Fig. 6E; Supplementary Fig. S6). In addition, specific STAT3 inhibitor S3I-201 abolished the effect of macrophages stimulated by oleic acid in the presence of heparanase on E0771 cells proliferation (Fig. 6E, dotted bars). Altogether, these findings further support the proposed involvement of heparanase in powering procancerous action of FFA-stimulated TAM in obesity-associated breast carcinoma and role of the IL6–STAT3 axis in this process.

Finally, to place our in vitro findings in the context of obesity-associated breast carcinoma in vivo, we compared the activation status of STAT3 in E0771 tumors growing in lean and obese, wt and heparanase-KO mice. As shown in Fig. 6F and G, immunostaining analysis revealed augmented STAT3 signaling (manifested by significantly increased levels of phospho-STAT3) in tumors growing in wt obese mice versus tumors growing in their lean littermates. Importantly, heparanase deficiency prevented obesity-induced STAT3 signaling, as no increase in STAT3 signaling was detected in tumors growing in Hpse-KO obese versus lean mice (Fig. 6F and G), echoing failure of obesity to accelerate E0771 breast carcinoma growth in Hpse-KO animals. Altogether, these data provide an explanation for the lack of tumor-accelerating effect of obesity in heparanase-deficient mice and highlight the previously unknown role of the enzyme in sustaining obesity–breast cancer link.

Obesity is a known risk factor in breast cancer (largely ER-positive; refs. 2–4, 6). Furthermore, once diagnosed, obese patients with breast carcinoma have worse clinical outcomes than their lean counterparts (3–5). Chronic, “smoldering” inflammation, adverse activation of macrophages, and altered production of estrogens are among key mechanisms through which excess adiposity fosters breast carcinoma progression and therapy resistance, acting both locally and systemically (3–6). Heparanase enzyme was recently linked to modulation of macrophage responses (32–35, 50) through TLR-dependent mechanism: intact cell surface HS is capable of suppressing TLR4 responses/macrophage activation, its degradation by heparanase abolishes this suppression (28), and cleavage of cell-surface HS by heparanase increases ligand accessibility of TLR4 (33). Moreover, heparanase-generated soluble HS degradation fragments stimulate TLR4 signaling in vitro (28, 30) and in vivo (31).

Here we show that heparanase is preferentially expressed in clinical and experimental obesity-associated breast carcinoma (Fig. 2) and orchestrates a complex multisite molecular program that occurs concurrently in adipose (Figs. 3C and 4) and breast tumor (Figs. 5 and 6F and G) tissues; involves tumor cells, hypertrophic “obese” adipocytes, and surrounding macrophages; and sustains obesity-accelerated breast carcinoma growth. Thus, our study not only defines heparanase as a marker of obesity-related, macrophage-driven breast carcinoma progression but also provides a mechanistic explanation for accelerated growth of estrogen-responsive breast tumors under obese state. In support of this notion, heparanase deficiency abolished obesity-accelerated progression of ER-positive E0771 tumor in vivo. Importantly, we found that heparanase deficiency did not abolish obesity-accelerated growth of estrogen-independent C57BL/6J-syngeneic Panc02 pancreatic carcinoma cell line (Supplementary Fig. S7). Panc02 is a well-established model of obesity-accelerated tumor growth in vivo (5, 7, 19) and represents one of the known 12 tumor types, other than ER-positive breast carcinoma, associated with obesity (1). Thus, our observations that heparanase deficiency abolishes obesity-accelerated progression of ER-positive E0771, but not of ER-negative Panc02 tumor model, further corroborate the mechanistic link between heparanase effect on tumor-promoting action of obesity and estrogen responsiveness.

It seems that macrophage-driven inflammatory events in adipose tissue, induction of heparanase in the tumor compartment, and breast carcinoma–promoting action of TAM are inexorably coupled in obesity-associated ER-positive breast cancer (as schematically presented in Fig. 7). The steps occurring in “obese” adipose tissue (Fig. 7, steps i–iv) involve adverse activation of atM, (supported by data depicted in Fig. 4A–C), that is, by saturated fatty acids locally released from the lipid-overloaded “obese” adipocytes (3, 9, 10), which are known to trigger TLR4. Activated atM secrete TNFα (Fig. 4A–C), a powerful trigger of aromatase expression in adipose fibroblasts (Fig. 7, step iii; refs. 9, 17, 18). Noteworthy, steps ii to iii (Fig. 7) require basic levels of heparanase present in the macrophages themselves, as heparanase deficiency renders macrophages less responsive to activation by saturated fatty acids (as in Fig. 4C) and abolishes their ability to induces aromatase expression in mouse obese adipose tissue in vivo and in human breast adipose fibroblasts ex vivo (as in Figs. 3C and 4D).

Induction of aromatase in adipose tissue eventually results in estrogen-mediated upregulation of heparanase in the breast carcinoma cells (Fig. 7 steps iv and v, supported by data depicted in Figs. 2A, C, D and 3A and B; also see refs. 23 and 25 for exact molecular mechanism through which estrogen induces heparanase in ER-positive breast carcinoma). During the following steps occurring in the breast tumor compartment, breast carcinoma cells that succeed in upregulating heparanase can effectively recruit/polarize TAM (Fig. 7, step vi, supported data on Fig. 5 and Supplementary Fig. S4), as increased levels of the enzyme enable acquisition of procancerous phenotype by macrophages stimulated by locally circulating FFA (supported by data depicted in Fig. 6 and in Supplementary Fig. S5 and S6). The above-described scenario results in increased rate of procancerous cytokines locally released by TAM within the tumor, activation of breast carcinoma-promoting signaling (i.e., STAT3), and accelerated tumor growth (Fig. 7, step vii, supported by findings depicted in Figs. 6B–D, F, G and 1C and 6E, in accordance). Thus, in obese state, presence of systemic circulating FFA (and perhaps other obesogenic compounds known to affect macrophage responses), combined with locally induced heparanase in ER-positive breast carcinoma tissue, creates “opportunity” for abnormal activation of TAM, whereas in the rest of tissues of the same patient (where heparanase is not expressed), levels of circulating FFA alone are not sufficient to trigger pathologic macrophage activation.

TAM are well-characterized target for tumor treatment (8) and specifically for uncoupling obesity–cancer link (5, 7, 19). Yet, translation of TAM-targeting approach into the clinic remains extremely challenging, in part, due to the highly heterogeneous and dynamic nature of TAM phenotypes (i.e., their ability to exert both pro- and antitumor activity, depending on different tumor settings; ref. 8). Thus, identification of heparanase as a novel molecular player that dictates the procancerous pattern of TAM activation in the obesity-related breast carcinoma can provide important information for translation of TAM-targeting approach into clinical practice and may also help better define patient populations in which this approach could be particularly beneficial. It should be noted that the limitation of this study is the lack of heparanase inhibition in vivo, which will be important for further translation of the basic mechanism revealed by our findings into therapeutic approach. In light of the above findings, the next question is whether the currently available heparanase inhibitors could disrupt the obesity–breast cancer linkage. Currently, several heparanase inhibitors, primarily HS-mimicking compounds such as pixatimod (PG545) and Roneparstat, undergo clinical testing (50). However, due to their structural similarity to soluble HS, the HS-mimicking compounds may also act as TLR ligands, stimulating macrophages regardless of their heparanase-inhibiting properties (32, 34). In addition, recently identified ability of heparanase to enhance NK-cell infiltration into tumors raised concerns that at least in some tumor types the potential adverse effect of heparanase inhibitors on NK cells may limit their antitumor effect (54). On the other hand, because antitumor function of NK cells was reported to be limited/lost under obese conditions (55), heparanase targeting is still relevant in obesity-associated breast carcinoma. Hence, ongoing development of heparanase-inhibiting compounds devoid of macrophage-activating properties (such as neutralizing antibodies, small-molecule inhibitors; ref. 50) will likely offer a better strategy to reduce breast carcinoma risk in an increasingly obese population and to suppress the breast cancer–promoting consequences of excess adiposity.

No potential conflicts of interest were disclosed.

Conception and design: M. Elkin

Development of methodology: E. Hermano, J.-P. Li, J. van der Vlag, T. Peretz, M. Elkin

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Goldberg, A. Sonnenblick, B. Maly, M.A.H. Bakker, I. Vlodavsky, T. Peretz, M. Elkin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Hermano, R. Goldberg, A.M. Rubinstein, A. Sonnenblick, D. Nahmias, M.A.H. Bakker, M. Elkin

Writing, review, and/or revision of the manuscript: R. Goldberg, A. Sonnenblick, J. van der Vlag, I. Vlodavsky, T. Peretz, M. Elkin

Study supervision: M. Elkin

Other (conduction of the experiments): E. Hermano, R. Goldberg, A.M. Rubinstein, D. Nahmias

Other (generation/characterization of clinical databases; review of the manuscript): T. Peretz

This study was supported by grants from the Israel Science Foundation (grant nos. 806/14, 1715/17), the Legacy Heritage Bio-Medical Program (grant no. 666/16), and the Mizutani Foundation for Glycoscience Research Grant, all awarded to M. Elkin. It was also supported by grants from the Israel Science Foundation (grant nos. 601/14, 2572/16) and the Israel Cancer Research Fund, awarded to I. Vlodavsky.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.