Background
Obesity and type II diabetes are associated with an increased risk of breast cancer and poorer clinical prognosis, particularly in postmenopausal women. Over the past several years, many studies have reported decreased breast cancer incidence and/or mortality in diabetics receiving the widely prescribed antidiabetic drug metformin relative to those receiving other diabetic drugs [
1‐
3]. Furthermore, these studies have found a dose-response relationship whereby women receiving the highest metformin dose for the longest duration show the most benefit [
4,
5]. Despite the routine use of this drug, the anticancer mechanisms of metformin are not well understood. Additionally, some studies have failed to show a beneficial effect of metformin as an antitumor agent [
6,
7]. Thus, there is ongoing interest in identifying patient populations who may benefit from metformin treatment, and the mechanisms by which metformin decreases cancer risk and/or improves tumor outcomes.
As a diabetic agent, metformin stabilizes glucose flux and reduces insulin resistance. It is used worldwide because of its low toxicity profile and low cost. Metformin activates AMP-dependent kinase (AMPK) to stimulate glucose uptake and glycogen synthesis, while suppressing gluconeogenesis, thereby improving whole-body insulin sensitivity. Beyond this, however, the exact mechanisms of action of metformin, even in diabetic patients, are not fully understood.
The antitumor activity of metformin has generally been attributed to its ability to decrease circulating insulin levels and improve whole-body metabolic health, as hyperinsulinemia has been associated with increased risk for breast cancer development, recurrence, and even death [
8,
9]. There is also a growing body of literature to suggest that metformin may directly target both tumor cells and tumor stem cells [
10‐
14]. These direct and indirect effects are not mutually exclusive, and it is possible that the effects of metformin involve a combination of the two.
Additional questions also remain regarding the patients and/or tumor subtypes for which metformin may be most effective. Preclinical studies suggest that at least some underlying whole-body metabolic dysfunction is needed to see beneficial effects of metformin, at least for breast cancer prevention [
15,
16]. Clinical data to date are primarily derived from diabetic patients receiving metformin and, from the few clinical trials completed to date, it is unclear which patients benefit most [
17‐
19]. The results of an ongoing clinical study are anticipated to shed light on this issue [
20]. In-vitro studies have demonstrated potential therapeutic utility against all breast cancer cell lines tested, but there is emerging preclinical and clinical data suggesting that metformin may be more effective in treating specific tumor subtypes (reviewed in [
21]). In cell lines, estrogen receptor (ER+) and tamoxifen-resistant breast cancers have been identified as a therapeutic target since metformin has been shown to inhibit the expression and function of ERα [
12,
13]. Consistent with these in-vitro studies, a recent clinical study reported that metformin specifically benefits women with ER+ as well as human epidermal growth factor receptor 2 (HER2)-positive breast tumors [
14]. Finally, we recently reported that tumor expression of organic cation transporter-2, which regulates metformin uptake, correlates with tumor responsiveness to metformin [
22]. Thus, evidence suggests both host- and tumor-specific targets, and while understanding of the anti-cancer mechanisms of metformin are emerging, there are still many questions that remain unanswered. Here we focus on host biology, namely the metabolic dysregulation that occurs during the menopausal transition, as a potential window of increased metformin efficacy.
In a rodent model of ER+ postmenopausal breast cancer we have shown that, during ovariectomy (OVX)-induced weight gain, an impaired ability to clear excess nutrients from the circulation and store them in mammary adipose tissue correlated with tumor glucose uptake, markers of proliferation, and tumor progression [
23]. Given these findings, we hypothesized that improving the whole-body metabolic response to excess calories during the window of weight gain that follows loss of ovarian function would decrease tumor growth and lead to improved tumor outcomes. We further hypothesized that both lean and obese animals would benefit from this treatment because they both experience overfeeding, rapid weight gain, and a decline in metabolic health in response to OVX [
23,
24].
Here, we report that metformin effectively decreased mammary tumor burden in both obese and lean animals, and also prevented the formation of new tumors in the postmenopausal period. In addition to improving overall metabolic health, metformin also decreased the number of aromatase-positive, CD68-positive macrophages within the tumor microenvironment, suggesting a new role for metformin in targeting immune cells. We therefore propose that the menopause and the peri-menopausal window represent a unique opportunity for metformin therapy, specifically for women with existing ER+ tumors and/or those who may be at risk for the development of postmenopausal breast cancer.
Methods
Animal care and treatment
Female Wistar rats (100–125 g, 5 weeks of age) were purchased from Charles River Laboratories (Wilmington, MA) and housed at the University of Colorado Anschutz Medical Campus Center for Comparative Medicine (22–24 °C; 12-h/12-h light-dark cycle) with free access to water. All procedures were approved by the Institutional Animal Care and Use Committee.
Animal model
Our OP-OR/OVX model of obesity and postmenopausal breast cancer was used as previously described [
25]. We and others have shown that tumors that develop using this method are similar to human breast tumors with regard to: 1) the percentage of tumors that are intraductal; 2) the progression of histologic stages from hyperplasia, to carcinoma in situ, to invasive cancer; and 3) steroid receptor status [
23,
25‐
27].
To induce obesity in these genetically susceptible rats, animals were individually housed in wire-bottomed metabolic cages to limit physical activity, and were given ad libitum access to a purified high-fat diet (HF; 46% kcal fat; Research Diets, New Brunswick, NJ; RD# D12344) for the duration of the study. Animals were ranked by their rate of weight gain from 5 to 15 weeks of age. Rats in the top and bottom tertiles of weight gain were matured to produce obese and lean animals, respectively. Rats from the middle tertile were removed from the study.
To induced mammary tumor formation, 55-day-old female rats (± 1 day) were given a single injection of the carcinogen 1-methyl-1-nitrosourea (MNU; 50 mg/kg; #ASI-701, Ash Stevens, Detroit, MI). Tumors were monitored by manual palpation at weekly intervals for the duration of the study and measured in three dimensions using digital calipers.
Body weight and food intake were monitored weekly, as previously described [
24,
28]. Body composition was determined on the day of OVX, 5 weeks post-OVX, and again at the time of sacrifice by quantitative magnetic resonance (qMR; EchoMRI Whole Body Composition Analyzer; Echo Medical Systems, Houston, TX).
In a rolling study design, animals were randomly assigned to either metformin treatment (2 mg/mL in the drinking water,
n = 7 lean and
n = 10 obese) or control group (water only,
n = 8 lean and
n = 9 obese) after at least one tumor in the animal reached a volume > 1 cm
3. Rats were maintained on their respective treatments for the duration of the study. This dose was chosen based on results of pilot studies that produced plasma metformin levels similar to those used clinically, and because this dose demonstrated antitumor efficacy in obese rats in our previous short-term studies, with no observable negative side effects [
22,
23]. Metformin treatment was initiated 1 week prior to OVX surgery to assure drug bioavailability during the critical window of rapid weight gain that immediately follows OVX. One week following the initiation of metformin treatment, the animals underwent surgical ovariectomy (OVX) under isoflurane anesthesia to mimic the postmenopausal state. At the time of OVX, biopsies of mammary tumors were obtained via fine needle aspiration (FNA).
Plasma measurements
Tail vein blood was collected on the second diestrus day of the estrous cycle [
24] during the week prior to OVX, at 5 weeks post-OVX, and again at the time of sacrifice. Blood was drawn during the latter part of the light cycle; plasma was isolated and stored at −80 °C until analyzed. Concentrations of insulin, leptin, amylin, and glucagon were simultaneously measured using the Rat Endocrine LINCOplex Kit 96 Well Plate Assay (RENDO-85 K; Millipore, St Charles, MO). Colorimetric assays were used to measure plasma free fatty acids (Wako Chemicals USA, Richmond, VA), glucose, triglycerides (TG), and total cholesterol (#TR15421, TR22321, and TR13521, respectively; Thermo Fisher Scientific, Waltham, MA).
Histological staining and imaging
Sections of formalin-fixed paraffin-embedded tissue (4 μm) were stained with hematoxylin and eosin (H&E) using a Sakura autostainer (Sakura Finetek, Torrance, CA). Mammary tumors were classified histologically by the criteria of Young and Hallowes [
29], and only adenocarcinomas were included in subsequent analyses. For immunohistochemical detection of progesterone (PR)-positive cells, 4 μm mammary tissue sections were stained with mouse monoclonal αPR, clone 6F11 at 1:100 (Vector Laboratories, Burlingame, CA). The dual localization of CD68 and aromatase was performed by staining for CD68 (Ab4059, Serotec, 1:200 dilution) sequentially followed by a mouse monoclonal aromatase antibody (clone 677) at a 1:100 dilution. The 677 monoclonal antibody was generated in one of the author’s laboratories (DPE) and has been validated extensively for specificity by immunohistochemistry (IHC) [
30‐
33]. For CD68, mouse on rat secondary antibody (MRT621H, Biocare) for 30 mins was used followed by 3,3′-diaminobenzidine chromogen (DAB; K3467, Dako, Carpinteria, CA). The slides were sequentially stained with aromatase primary followed by Rat on mouse AP polymer (MALP521, Biocare) followed by Permanent red chromogen (K0640, Dako, Carpinteria, CA). The adipophilin primary antibody (LS-C348703, Lifespan Biosciences) was incubated on rat livers at 1:300 dilution for 60 min followed by mouse on rat secondary antibody (MRT621H, Biocare) for 30 min and DAB chromogen. Livers from lean and obese pre-OVX animals were obtained from a separate cohort of rats in which animals were terminated with the ovaries intact. All slides were counterstained with hematoxylin (S330130, Dako, Carpinteria, CA).
For mammary tumor ER and PR analysis, at least 11 tumors per group, and 8–10 fields/section (40× objective) were evaluated. Given the effectiveness of metformin in shrinking tumors in this study, the number of samples from metformin-treated rats with sufficient tissue to perform IHC analysis was limited; thus, for the CD68 and aromatase IHC in tumors and adjacent tumor border, our analysis was restricted to tumors from obese control and metformin-treated rats. CD68 and aromatase stained slides were scanned using an Aperio Scanscope3 system (Aperio, Vista, CA) at 20× magnification, corresponding to 0.43 μm per pixel which enables high-resolution access to the entire tissue section via a virtual image. Images were evaluated using Imagescope software and the signal captured and quantitated using Aperio algorithms. For tumor border analysis, the tumor boundary was outlined on H&E stained slides using the Aperio Annotation tool within Leica Image Scope (Leica Technologies, CA) by an MD pathologist (SJ). The tumor boundary images were then exported and overlaid (imported) onto an adjacent serial section dual stained for CD68 and aromatase using the Export/Import tool within Leica Image Scope. The tumor border was then defined as 100 μm external to the tumor boundary and captured using the Aperio Ruler tool. The Aperio system was also used for liver adipophilin quantitation, where a minimum of nine livers per group were evaluated.
In-vitro macrophage differentiation
Rat macrophages were derived from a pooled bone marrow stock aspirated from femurs and tibias of 7-week-old female Wistar rats. Once isolated, marrow was cultured in vitro in Dulbecco’s Modified Eagle’s medium (DMEM) low glucose with 30% L929-cell conditioned media (as a source of macrophage colony-stimulating factor (M-CSF)) containing either 5 ng/mL lipopolysaccharide and 12 ng/mL interferon (IFN)-gamma to promote differentiation to an M1 phenotype, or 10 ng/mL interleukin (IL)-4 to promote an M2 phenotype. Following 48 h in differentiation medium, cells were rinsed twice with phosphate-buffered saline (PBS) and harvested in cell lysis buffer (50 mM Tris pH 7.4, 150 mM NaCl, 2.0 mM EDTA, 50 mM NaF, 5.0 mM sodium orthovanadate, 1% Triton X-100, 1% deoxycholate, 0.1% SDS) supplemented with Halt™ Protease and Phosphatase Inhibitor cocktail (Thermo Scientific). Lysates were centrifuged for 20 min at 14,000 g, and total protein concentration of the supernatant was determined by BioRad protein assay according to the manufacturer’s instructions.
ProteinSimple© Western blotting and analysis
Protein levels of aromatase in M1 and M2 macrophages were measured using the Simple Western size-based capillary electrophoresis system (WES, ProteinSimple, San Jose, CA). Two different anti-aromatase primary antibodies were used, including Novus NB100-1596 (1:50 dilution) and clone #677 used for the previously described IHC (1:25 dilution). All procedures were performed according to the manufacturer’s protocol and immunodetection was conducted with default settings. Data were analyzed with ProteinSimple Compass software. Additional controls included the use of a blocking peptide against aromatase antibody #1 (Novus NB100-1596PEP) to verify the band size and specificity using this antibody on the WES system, and the use of forskolin-treated steroidogenic human granulosa-like tumor cells (KGN cells) to induced aromatase expression as a positive control with aromatase antibody #2.
Statistical analysis
Data were examined with SPSS 24.0 software by ANOVA or χ2 analysis for nominal and ordinal data, respectively. Relationships between variables were assessed with the Spearman correlation coefficient. In some cases, data were analyzed by analysis of covariance with a specified covariate in the model.
Discussion
Menopause represents a lifecycle window of breast cancer risk that may be highly amenable to interventions that decrease risk. During menopause, energy balance, circulating hormones, chemokines and cytokines, and body fat distribution are in flux, and this is also the critical time when the tumor-promoting effects of obesity emerge [
53‐
55]. Thus, interventions targeting the metabolic flux of menopause may effectively reduce breast cancer incidence and/or lethality. Using a rat model of postmenopausal breast cancer, our goal was to determine if targeting metformin treatment to the window of ‘menopause’-induced weight gain could decrease tumor growth and improve tumor outcomes. Similar to menopause in women, OVX in this model induced weight gain and increased adiposity in all animals, regardless of their lean/obese status prior to OVX. Weight gain was associated with a decline in metabolic health, as demonstrated by increased liver fat deposition and adipose-tissue inflammation. Specifically, we identified a subtype of aromatase-positive, M2-like macrophages to be elevated in mammary adipose tissue post-OVX. Within this context of OVX-induced metabolic dysfunction, metformin decreased the size of existing tumors and prevented formation of new tumors. The antitumor effects of metformin were associated with a decrease in adipose inflammation, measured by a reduction in the number of aromatase expressing CD68
+ macrophages. Overall, our data suggest that a subtype of aromatase-positive, M2-like macrophages are elevated with OVX-induced weight gain, providing a growth advantage to ER
+ tumors in the absence of ovarian hormones. These macrophages are targeted by treatment with metformin, possibly mitigating the protumorigenic effects of OVX-induced weight gain through estrogen deprivation. While nonaromatase-dependent mechanisms of metformin most certainly also contribute to tumor reduction in this model, this newly identified mechanism of action warrants further investigation.
This study further supports our ‘dual requirement’ hypothesis of obesity and postmenopausal breast cancer. Our early work in this model established that both impaired metabolic regulation that underlies obesity and a positive energy imbalance are required for the emergence of obesity-associated tumor promotion after menopause. This combination of impaired metabolism and postmenopausal weight gain has direct effects on mammary tumors, specifically increasing tumor expression of PR, promoting a glycolytic/lipogenic gene expression profile, and promoting tumor glucose uptake [
23]. More recently, we have also shown that this combination of obesity and OVX-induced overfeeding leads to nuclear localization of the androgen receptor, which promotes the growth of ER+ tumors under conditions of low estrogen availability after OVX [
35]. Importantly, in that study we found a role for the inflammatory cytokine IL-6 in sensitizing breast cancer cells to low testosterone levels. Our current study now extends this work to suggest that rapid weight gain following OVX/menopause is associated with increased aromatase expression in mammary macrophages. Work in endometrial cancer has demonstrated that tumor cell production of IL-6 leads to upregulation of aromatase in stromal cells, creating a cycle that drives tumor proliferation [
56]. A similar IL-6-mediated increase in aromatase gene expression has been demonstrated in murine macrophages [
57]. The existence of a similar paracrine mechanism in the context of postmenopausal breast cancer warrants further investigation.
The combination of impaired metabolic health and menopause-induced weight gain were likely critical to the anticancer effects of metformin observed in this study. A review of the literature would suggest that metformin is most effective when one or more of the following are present: 1) consumption of a moderate to high-fat diet; 2) poor metabolic health (insulin insensitivity, metabolic disease, etc.); 3) weight gain; and/or 4) increased adiposity (overweight or obesity). Our previous work demonstrating beneficial effects of metformin have all been conducted in animals consuming a high-fat diet [
22,
23]. There are several examples where metformin had minimal or no impact on mammary tumor outcomes in the context of a low-fat diet [
15,
58,
59]; however, in studies where medium [
60] or high-fat [
61] diets were used, metformin improved tumor outcome. Beneficial effects of metformin have also been reported in a study where 5% sucrose was added to the water of the animals [
62], suggesting that this may have been sufficient to impair the metabolic health of these animals to an extent where the effects of metformin could be realized.
While the role of tumor-associated macrophages in breast cancer development and progression has been studied extensively over the past decade (reviewed in [
63]), there are only a limited number of studies that have focused specifically on macrophage production of aromatase. Using IHC staining on serial sections, Mor and colleagues [
46] demonstrated the presence of CD68
+, aromatase-positive macrophages both around and within human breast cancers. Using in vitro assays, they extended this work to show that aromatase expression and activity is acquired by tissue-activated macrophages but not by their circulating monocyte precursors. Furthermore, conditioned medium from activated macrophages was sufficient to stimulate the growth of estrogen-responsive MCF-7 cells—an effect blocked by the aromatase inhibitor letrazol. This demonstrates that, at least in vitro, macrophages can produce sufficient levels of estrogens to stimulate the growth of estrogen-responsive breast cancer cells. Our findings are the first, to our knowledge, to identify aromatase expression as a feature of a subpopulation of protumorigenic M2-like mammary macrophages that arise in the context of obesity. We speculate that, in our in-vivo model, locally produced estrogens reach sufficient levels to activate ER in mammary tumors, and the ability for metformin to decrease this local production of estrogen contributes to its antitumor effects. The question whether inflammatory cytokines such as IL-6 cooperate with stromal-derived estrogen and sensitize breast cancer cells to ER, as we observe for testosterone signaling through the androgen receptor [
35], remains to be determined.
Our data build upon the work of Dannenberg’s group who have demonstrated a causal link between obesity-induced inflammation and aromatase expression in the mammary gland. Their work shows that, in obese mice, release of free fatty acids from adipocytes activates NF-kB in the stromal vascular fraction of adipose tissue, which increases proinflammatory cytokine production [
45]. In cell culture models, they have demonstrated that that these proinflammatory mediators (tumor necrosis factor (TNF)α, IL-1β, and prostaglandin E
2(PGE
2)) produced by cells in the stromal fraction of mammary glands from obese mice stimulate aromatase in preadipocytes [
45]. They have further extended these findings to demonstrate increased inflammation, aromatase expression, and aromatase activity in the breast of overweight and obese women [
42] and in a subset of nonoverweight women (body mass index (BMI) < 25 kg/m
2) who had underlying systemic metabolic dysfunction [
41]. Our work extends these pioneering studies and indicates that M2-like macrophages themselves can produce aromatase. In our rat model, mature mammary adipocytes do not appear aromatase-positive, and estradiol in rat mammary adipose tissue was below the level of detection by mass spectrometry [
35]. It is possible that the role of adipocytes in aromatase production could be model- or context-dependent. However, combined, these data highlight the fact that many stromal cell populations may contribute to local aromatase production under different conditions.
Acknowledgements
We would like to thank J. Higgins and M. Jackman for their helpful discussions throughout this study. We are also grateful for the technical assistance provided by S. Edgerton, K. Hedman, G. Johnson, D. Landrock, A. Lewis, J. Lopez, C. Mahan, R. Oljira, and V. Wessells.