Normalizing glucose levels reconfigures the mammary tumor immune and metabolic microenvironment and decreases metastatic seeding

Abstract

Obesity and diabetes cumulatively create a distinct systemic metabolic pathophysiological syndrome that predisposes patients to several diseases including breast cancer. Moreover, diabetic and obese women with breast cancer show a significant increase in mortality compared to non-obese and/or non-diabetic women. We hypothesized that these metabolic conditions incite an aggressive tumor phenotype by way of impacting tumor cell-autonomous and tumor cell non-autonomous events. In this study, we established a type 2 diabetic mouse model of triple-negative mammary carcinoma and investigated the effect of a glucose lowering therapy, metformin, on the overall tumor characteristics and immune/metabolic microenvironment. Diabetic mice exhibited larger mammary tumors that had increased adiposity with high levels of O-GlcNAc protein post-translational modification. These tumors also presented with a distinct stromal profile characterized by altered collagen architecture, increased infiltration by tumor-permissive M2 macrophages, and early metastatic seeding compared to non-diabetic/lean mice. Metformin treatment of the diabetic/obese mice effectively normalized glucose levels, reconfigured the mammary tumor milieu, and decreased metastatic seeding. Our results highlight the impact of two metabolic complications of obesity and diabetes on tumor cell attributes and showcase metformin’s ability to revert tumor cell and stromal changes induced by an obese and diabetic host environment.

Introduction

Type 2 diabetes mellitus (T2DM) and obesity are considered worldwide epidemics by the World Health Organization (WHO) due to their high prevalence and increasing incidence [1]. Obesity is one of the most significant risk factors for type 2 diabetes [2] and not only impacts the incidence of diabetes, but it also exacerbates the risk of developing diabetes-related complications [1]. The prevalence of obesity in the US was 42.4% between 2017–2018 [3]. Researchers predict that by 2030, nearly half of the adult population will be obese, with around 25% suffering from severe obesity [4]. Diabetes, another prevelant condition, affects around 10.5% of the adult population in the United States, while 34.5% are considered prediabetic [5]. Both obesity and diabetes create distinct systemic metabolic pathophysiological syndromes that predispose women to several diseases, including breast cancer [6]. Moreover, obese women with breast cancer show a 33% increase in mortality compared to non obese patients [7].

Hyperglycemia is a shared manifestation in patients with both type 1 (insulin dependent) and type 2 (insulin independent) diabetes, but the underlying pathological and metabolic mechanisms are different between the two diabetic conditions. Type 1 diabetes is a condition of insulin deficiency due to autoimmune destruction of the pancreatic beta cells, unlike T2DM, that is characterized with insulin resistance, and generally associated with metabolic syndromes and obesity. This might explain why cancers linked to type 2 diabetes are distinct from cancers associated with type 1 diabetes [6, 8]. Individuals with type 2 diabetes but not type 1 diabetes, manifest with an increased risk of developing breast cancer [8, 9]; furthermore, type 2 diabetes is associated with advanced breast cancer stage and higher mortality [10]. Metformin (1,1-dimethyl biguanide) is a widely used first line antihyperglycemic treatment for patients with type 2 diabetes. The beneficial effects of metformin are attributed to lowering blood glucose by inhibiting hepatic gluconeogenesis and enhancing insulin sensitivity [11]. In addition to its glucose lowering effect, studies have reported its anti-tumor effects which are attributed to multiple mechanisms [11]. These include inhibition of mitochondrial complex I thereby limiting ATP production which in turn activates ATP-sensing AMPK, leading to cessation of growth. Interestingly, AMPK activation tempers cellular O-GlcNAcylation [12, 13], a protein posttranslational modification that is robustly upregulated in diabetes. Studies have suggested a possible association between the use of metformin and decreased risk of cancer incidence and mortality [14–21]. Independent of AMPK, metformin inhibits cancer cell proliferation by suppressing the production of mitochondrial-dependent metabolic intermediates required for cell growth [22].

In this study, we sought to understand how an obese and diabetic host impacts tumor cell-intrinsic and tumor cell-extrinsic attributes. We utilized a type 2 diabetic/obese mouse model to examine the effect of the host metabolic milieu on growth and metastatic behavior of mammary tumors. Diabetic/obese mice exhibited larger mammary tumors and frequent metastases compared to non-diabetic/lean or metformin-treated diabetic/obese mice. In addition to altering the tumor-intrinsic characteristics, the host state of obesity and hyperglycemia fostered a tumor-favoring macrophage immune microenvironment, that was tempered by metformin.

Methods and Materials

Analysis of data from breast cancer patients

We used the UAB institutional database informatics resource for integrating biology and the bedside (i2b2) to extract data on female patients with a malignant breast cancer diagnosis in the University of Alabama at Birmingham (UAB) Health System. We extracted demographic data (age, race/ethnicity, and marital status), vital status, Body Mass Index (BMI), and date of death as confirmed in patient’s medical record. Malignant breast cancer and T2DM diagnoses were confirmed using International Classification of Diseases, ninth and tenth revision (ICD-9, ICD-10), or Systematized Nomenclature of Medicine Clinical Terms (SNOMED) diagnosis codes. Survival probabilities were estimated using Kaplan Meier and cumulative hazard estimates. Multivariable analyses were performed using Cox Proportional Hazard models to estimate hazard ratios (HRs) for diabetes in relation to all-cause mortality. Models were adjusted for age, race/ethnicity, marital status, and body mass index. For determining relevance with respect to age, we applied the median age of our population (67 years) as the cut-off (interquartile range = 18 years).

Cell Culture

Luciferase-expressing E0771 murine breast cancer cell line was obtained from ATCC (Manassas, VA) and cultured in Roswell Park Memorial Institute (RPMI) 1640 Medium (D-Glucose 2 gm/L) (Life Technologies; Carlsbad, CA) supplemented with 10% heat-inactivated fetal bovine serum (FBS) (Gibco, Life Technologies Carlsbad, CA) under puromycin (Millipore Sigma, Burlington, MA) (1mg/ml) selection. Cells were maintained at 37°C in a humidified 5% CO₂ incubator.

Animal study

The animal studies have been conducted in accordance with the Institutional Animal Care and Use Committee (IACUC) of UAB. C57BL/6J female mice were obtained from Jackson Labs (Bar Harbor, ME), Mice were fed with either high fat diet (Rodent diet with 60 kcal% Fat # D12492ji, Research Diets, New Brunswick, NJ) or matching control diet (Rodent diet with 10 kcal% Fat #D12450ji) for 72 days. Food was changed twice a week. Streptozotocin (STZ) (Enzo Life Sciences # ALX-380–010-G001 Farmingdale, NY) dissolved in sodium citrate 2.54% (pH 4.5) was injected intraperitoneally (IP) at a very low dose of (40mg/kg/day) on days 15–17 and 43–45 to stress but not kill pancreatic β-cells. This results in a more T2D-like phenotype [23]. Mice were fasted four hours before injection. STZ was prepared fresh daily at a concentration of 7.5 mg/mL and injected within 10 minutes of preparation. Metformin (Millipore Sigma, Burlington, MA; CAT# D150959) was suspended in deionized water (250 mg/kg/day) and administered daily by oral gavage starting at day 47 from initiation of HFD. Body weight and blood glucose were measured once weekly. Blood glucose was measured using CONTOUR®NEXT ONE system (Ascensia Diabetes Care, Parsippany, NJ).

For tumor formation, luciferase-expressing E0771 cells (1 × 10⁵), suspended in Hank’s Balanced Salt Solution (Life Technologies, CA, USA), were injected into the inguinal mammary fat pad (MFP) of female C57BL/6J mice at day 54 from initiation of the High fat diet (HFD). Tumor progression was documented by caliper measurements two times weekly and by bioluminescent imaging (BLI) using the IVIS Lumina III Living Image system (Perkin Elmer, Waltham, MA) once weekly. Briefly, mice were injected intraperitoneally with D-luciferin (150 mg/kg body weight) and anesthetized using isoflurane gas. Ten minutes later, images were obtained and analyzed using IVIS Lumina III Living Image software (Perkin Elmer).

Glucose tolerance test (GTT)

Mice were fasted for six hours beginning at 12 AM; baseline blood glucose was measured, and then mice were injected IP with D-glucose (1g/kg, Millipore Sigma D7528, Burlington, MA). Serial blood samples were collected from the tail vein at 15, 30, 45, 60, 90, and 120 minutes after glucose administration. Area under the curve (AUC) and statistical significance were analyzed using GraphPad Prism version 8.

Hematoxylin and Eosin (H&E) staining

Mouse tumors and tissues were fixed in formalin and embedded in paraffin. Sections (5μm) from paraffin-embedded tissue were deparaffinized and rehydrated through passages in xylene and gradients of ethanol. Slides were then immersed in Harris Hematoxylin (diluted 1:4 in tap water) for 2 minutes, de-stained by dipping in acid fast ethanol for 10 times, rinsed with tap water, immersed in Eosin Y (Leica Biosystems, Wetzlar, Germany) for 3 minutes, and rinsed again with tab water. Tissues were dehydrated and the slides were mounted in Cytoseal (ThermoScientific, Waltham, MA) with coverslips.

Immunostaining

Mouse tumors and tissues were fixed in formalin and embedded in paraffin. Sections (5 μm) from paraffin-embedded tissue were deparaffinized and rehydrated through passages in xylene and gradients of ethanol. Heat-induced antigen retrieval was performed in boiling sodium citrate buffer for five minutes, followed by incubation with Dual Endogenous Enzyme Block (Dako, Carpinteria, CA) for 15 minutes. Following a 5-minute wash with Tris-buffered NaCl solution supplemented with 0.01% Triton- X-100, tissues were incubated with the primary antibody PCNA (PC10; Cell Signaling #2586, Danvers, MA) 1:1000 overnight at 4°C. To stain tissue for overall O-GlcNAcylation, samples were blocked with 3% goat serum for 40 minutes and then incubated with RL2 Antibody (1:4000 Abcam Cambridge, UK ab201995) overnight at 4°C. After washing, EnVision+ System- HRP Labelled Polymer Anti-Rabbit (Dako) was applied for 40 minutes at room temperature. For signal visualization, tissues were incubated with Liquid DAB+ Substrate Chromogen System (Dako) for 3 minutes and counterstained with Harris Hematoxylin (diluted 1:4 in tap water). After tissue dehydration, slides were mounted in Cytoseal (ThermoScientific, Waltham, MA) with coverslips. Images were visualized by Nikon Eclipse E200 and captured using the 40x objective of the DS-L4 microscope system (Nikon). PCNA positive nuclei were evaluated using Fiji Image-J software [24]. For evaluation of RL2 staining, the intensity of staining of tumor cells was assessed as 0 (no staining) to 4 (strongest possible intensity of staining). The immunoscore was derived as the product of the percentage of cells at each intensity and the corresponding intensity. The products were added to get an immunoscore for the section [25].

CD80 (Abcam ab254579 1:500) / CD68 (Santa Cruz SC-20060 1:50) double IHC staining was performed using multiview plus IHC kit (ADI-950–100, Enzo Life Science, Farmingdale, NY) according to manufacturer’s protocol. For isolectin staining, slides were deparaffinized and rehydrated and blocked for dual peroxidase for 15 minutes. Next, the slides were incubated in blocking buffer with 0.1% BSA for 20 minutes followed by incubation with primary antibody [Isolectin B4, Biotinylated (Vector Labs B-1205, Burlingame, CA)] 1:100 for 1 hour, followed by incubation with [Vectistain] elite ABC reagent according to manufacturer’s protocol (Vector Labs PK-6100). For signal visualization, tissues were incubated with Liquid DAB+ Substrate Chromogen System (Dako) for 4 minutes and counterstained with Harris Hematoxylin (diluted 1:4 in tap water). After tissue dehydration, slides were mounted in Cytoseal (ThermoScientific) with coverslips.

For immunofluorescent histology staining, paraffin-embedded tissues were deparaffinized and rehydrated through passages in xylene and gradients of ethanol. Heat-induced antigen retrieval was performed in sodium citrate buffer for five minutes, followed by blocking in PBS+10%BSA blocking for 1 hour. Tissues were then incubated with the primary antibody CD206 (ab64693) 1:500 for 1 hour at 4°C followed by washing and one hour incubation with Alexa Fluor 488 secondary anti-Rabbit antibody (Life Technologies) 1:500. In order to enhance the signal we used Vector TrueVIEW Autofluorescence Quenching Kit with DAPI (Vector Labs) according to manufacturer’s protocol.

Picrosirius red staining

Slides were deparaffinized and rehydrated, and stained using Picrosirius red stain kit (Polysciences CAT# 24901, Warrington, PA) according to manufacturer’s protocol. Images were captured using a Nikon A1R HD Confocal Microscope (Nikon) using 40X lens under bright field and polarized light.

Apoptosis

Apoptosis was assessed by Click-it TUNEL colorimetric IHC detection system (Invitrogen #C10625, Carlsbad, CA) according to manufacturer’s protocol. TUNEL positive nuclei were evaluated using Fiji Image-J software [24].

Statistical analysis

As suitable, an unpaired t-test or one-way or two-way ANOVA was applied to analysis using GraphPad Prism version 8. Comparisons were considered statistically significant for p-value < 0.05.

Results

Obesity and diabetes negatively affect survival outcomes of breast cancer patients

In order to investigate how diabetes and obesity impact breast cancer survival outcomes, we used the i2b2 database to extract data on female patients with a malignant breast cancer diagnosis. The analytic cohort consisted of 11,241 female patients with malignant breast cancer diagnosed and treated at the UAB Health System between 2003 and 2020. A total of 1,972 (17.5%) patients had a diagnosis of T2DM and 8,848 (78.7%) were overweight, obese, or morbidly obese (BMI ≥ 25) with 5,992 (53.3%) being obese or morbidly obese (BMI≥30) (Table 1). The ten-year survival probability was higher in non-diabetic vs diabetic breast cancer patients (93.2% vs 85.5%) and in non-obese vs obese (92.6% vs 90.4%) (Figure 1A). The differences in survival probabilities were statistically significant in both groups (Table 2). In breast cancer patients with T2DM, the mortality rate was higher compared to individuals without diabetes (hazard ratio (HR) = 1.52, 95%CI 1.25 – 1.85, p-value < 0.001) adjusted for age, race/ethnicity, marital status, and BMI (Figure 1B, Supp. Table 1). Black or African American women presented with a greater mortality rate compared to White women. Quite unexpectedly, the data revealed that women who are not married have worse outcomes compared to married women – while the reasons for this are being studied by groups involved in implementation science, it is likely that better economic resources and greater social support may influence the treatment decisions, socioeconomic status, and stress-related variables [26]. As such, conditions of obesity and diabetes negatively impact the survival outcomes in breast cancer patients.

Table 1:

Demographic characteristics of female breast cancer patients with diabetes mellitus or obesity

Characteristic	All Breast Cancer n = 11,241	Diabetes Mellitus n = 1,972	BMI ≥ 25 n = 8,848	BMI ≥ 30 n = 5,992
Age at time of presentation
Mean years (SD)	66.1 (±13.5)	70.2 (±11.6)	66.1 (±13.0)	65.6 (±13.1)
Median (range)	67 (34 – 96)	70 (41 – 96)	67 (34 −95)	66 (34 – 94)
Race
Non-Hispanic White	7,905 (70.3%)	994 (50.4%)	6,072 (68.6%)	3,802 (63.5%)
Black	2,497 (22.2%)	870 (44.1%)	2,256 (25.5%)	1,833 (30.6%)
Other	403 (3.6%)	86 (4.4%)	298 (3.4%)	215 (3.6%)
Unknown	436 (3.9%)	22 (1.1%)	222 (2.5%)	142 (2.4%)
Ethnicity
Hispanic	9,961 (88.6%)	1,902 (96.5%)	8,225 (93.0%)	5,625 (93.9%)
Non-Hispanic	102 (0.9%)	15 (0.8%)	84 (0.9%)	51 (0.8%)
Unknown	1,178 (10.5%)	55 (2.8%)	539 (6.1%)	316 (5.3%)
Marital Status
Married	6,416 (57.1%)	850 (43.1%)	4,978 (56.3%)	3,185 (53.1%)
Other	4,556 (40.5%)	1,113 (56.4%)	3,756 (42.4%)	2,736 (45.7%)
Unknown	269 (2.4%)	9 (0.5%)	114 (1.3%)	71 (1.2%)
Vital Status
Alive	10,210 (90.8%)	1,673 (84.8%)	7,989 (90.3%)	5,401 (90.1%)
Dead	1,031 (9.2%)	299 (15.2%)	859 (9.7%)	591 (9.9%)

Figure 1:

A. Kaplan Meier overall survival curve for diabetic vs non-diabetic breast cancer patients from i2b2 database. Differences in survival probabilities were statistically significant in both groups; Log rank test p < 0.001

B. Nelson–Aalen cumulative hazard estimates for diabetic vs non-diabetic breast cancer patients from i2b2 database. Differences in survival probabilities were statistically significant in both groups p < 0.001

Table 2:

Survival probabilities for female breast cancer patients with diabetes mellitus or obesity

	5-year survival probability (95% CI)	10-year survival probability (95% CI)	Log rank test P-value
Diabetes
No	95.3 (94.8 – 95.8)	93.2 (91.8 – 94.4)	<0.001
Yes	90.4 (88.8 – 91.8)	85.5 (81.2 – 88.9)
BMI ≥ 25
NO	95.5 (94.3 – 96.5)	92.0 (85.8 – 95.6)	0.016
Yes	93.9 (93.3 −94.4)	91.2 (89.7 – 92.6)
BMI ≥ 30
NO	94.8 (94.0 – 95.5)	92.6 (90.3 – 94.4)	0.01
Yes	93.6 (92.9 – 94.3)	90.4 (88.2 – 92.2)
Diabetes and BMI ≥ 30
NO	95.4 (94.7 – 96.1)	93.3 (90.8 – 95.1)	<0.001
Yes	90.1 (89.2 – 92.5)	85.6 (80.5 – 89.5)

High-fat diet / low-dose streptozotocin-induces a state of type 2 diabetes in mice that is attenuated by metformin

In order to study the effect of obesity-associated type 2 diabetes on breast cancer, we utilized a mouse mammary tumor model that was nutritionally induced for obesity and diabetes [23, 27]. This model also presented us with the opportunity to evaluate the effectiveness of metformin, a widely used anti-diabetic drug [11], on tumor growth and progression. We initiated this study by first feeding the mice a nutrient-dense diet (HFD). As detailed in Figure 2A, we administered low dose streptozotocin (40 mg/kg) to induce type 2 diabetes in C57BL/6J mice. Mice fed a HFD developed a consistent and significant increase in body weight compared to control diet-fed and metformin-treated mice (2-way ANOVA, p < 0.0001) (Figure 2B; Supplementary Figure 1A-B). As a well-accepted morphometric measure of obesity, we measured the ratio of body weight (BW):tibia length (TL). Expectedly, this ratio was significantly higher in mice fed a HFD compared to control diet (p = 0.01) (Supplementary Figure 1C). T2DM is associated with cardiac hypertrophy as assessed by heart weight (WH):TL [23] [28]; however, we did not identify differences in this measurement (Supplementary Figure 1D), likely because our assessment spanned a much shorter interval than is used for cardiac-related investigations [29]. Six weeks after this regimen, but prior to metformin administration, the average blood glucose levels for HFD/STZ group were significantly higher (219±70.5 mg/dl) compared to the control group (Lean) (147±26.1 mg/dL) (p=0.03). A cohort of diabetic mice were treated with metformin (Ob/D+Met; 250 mg/kg p.o) as outlined in Figure 2A. Metformin significantly reduced the average blood glucose beginning at week 7 [167.4±19.4 mg/dL in Ob/D+Met mice compared to 234±33.1 mg/dL in untreated obese and diabetic (Ob/D) mice (p=0.04)] (Figure 2C). Thereafter, the glucose levels were sustained at a steady lower level. Type 2 diabetes is characterized by glucose intolerance. The glucose tolerance test (GTT) is widely utilized in research and clinical practice to detect individuals with normal or impaired glucose tolerance and individuals with type 2 diabetes [30]. To quantify the extent of diabetes, we performed an intraperitoneal GTT. Ob/D mice showed significantly lower glucose tolerance (2-way ANOVA, p<0.0001) compared to mice fed a control diet, providing evidence that the mice developed glucose intolerance characteristic of type 2 diabetes. Furthermore metformin-treated mice showed significantly improved glucose tolerance compared to untreated Ob/D mice (p < 0.0001) (Figure 2D). Additionally, AUC_Glucosewas significantly higher in Ob/D mice compared to mice on a control diet (p=0.0002), and this effect was mitigated in metformin treated mice (one-way ANOVA, p=0.04) (Figure 2E, Supplementary Figures 1E, 1F). Thus, metformin alleviates the state of type 2 diabetes that is invoked by a regimen of high-fat diet and low-dose streptozotocin.

Figure 2: High-fat diet/low-dose streptozotocin-induces a state of type 2 diabetes in mice that is mitigated by metformin.

A. Schematic for in vivo T2DM induction and metformin treatment. B. Average body weight (g) weekly measurements through the course of experiment. The difference is statistically significant between groups (2-way ANOVA, p = 0.0001). C. Blood glucose (mg/dl) weekly measurements through the course of experiment; red arrow represents the start of metformin at week 6. The difference was statistically significant (2-way ANOVA, p = 0.0002). D. Blood glucose measurements (mg/dl) during the intraperitoneal glucose tolerance test (1 g/kg) following fasting for 0, 15, 30, 45, 60, 90 & 120 min. Difference was statistically significant (2-way ANOVA, =<p = 0.0001). E. Area under the curve for glucose was calculated using the Receiver Operator Characteristic curve (ROC). Results are represented as mean ± SD (n = 5–7) (one-way ANOVA, p = 0.04).

Blood glucose levels correlate with mammary tumor growth

To investigate the effect of an obese and diabetic host on breast cancer, we injected mice with luciferase expressing E0771 murine mammary carcinoma cells into the right inguinal mammary fat pad on day 54 from the initiation of diet administration, having ascertained that the HFD-fed mice had elevated blood glucose levels. We monitored tumor growth by caliper measurement and BLI. Ob/D mice showed a significantly higher luciferase signal in the tumors compared to the lean mice and Ob/D+Met mice (2-way ANOVA, p < 0.03) (Figures 3A, B) This difference was more evident at day 19 following injection (one-way ANOVA, p <0.045) (Figure 3C), where reduced blood glucose levels in metformin-treated mice correlated with a remarkably reduced luciferase signal from the tumors. Moreover, at day 20 following tumor cell injection, Ob/D mice supported larger tumors compared to lean mice and Ob/D+Met mice (mean tumor volumes 740.5±438 mm³, 536.3±323.5 mm³, and 436.2±234.8 mm³ respectively), although these differences did not reach statistical significance (Figure 3D). In addition, the luminescence signal suggested a modest, albeit significant positive correlation (r=0.24, p=0.02) with blood glucose measures (Figure 3E). These findings taken together illustrate that obesity and the overall state of hyperglycemia promotes mammary cancer growth and metformin abrogated this effect.

Figure 3: Blood glucose levels correlate with mammary tumor growth.

A. Representative bio-luminescence images of mice with mammary tumors at D19 of E0771-FLUC tumors (n = 7 in each group). The color scale represents radiance (Photons/sec/cm³). B. Luminescence total flux (photons/sec) for the mammary fat pad tumors measured at D7, D14 D19 after injection of E0771 cells. Difference is statistically significant (2-way ANOVA, p= 0.03) C. Total Flux (p/s) between the groups at D19 after injection (n=7 per group). Results shown are mean ± SD (One-way ANOVA p=0.04) (unpaired t-test lean vs Ob/D p=0.04, Ob/D vs Ob/D+Met p=0.07). D. Mean tumor volume (mm³) = (length × width²) / 2 at Days 12, 15, 18 & 20 from injection; results are depicted as mean ± SD (n=7 each group) (2-way ANOVA p=ns). E. Graph depicting correlation between luminescence signal (p/s) at D19 and blood glucose measurement (mg/dl) at D14 (R square=0.49) (p=0.02).

Host obesity and diabetes alter tumor O-GlcNAcylation and adiposity

O-linked β-N-acetylglucosamine (O-GlcNAc) is a sugar attachment to the side chain hydroxyl of serine or threonine residues on proteins.Functioning as a central communicator of nutritional status, O-GlcNAcylation controls key signaling and biological processes such as signal transduction, transcription, cell cycle progression, and metabolism. O-GlcNAc signaling is intertwined with cellular metabolism; the donor sugar for O-GlcNAcylation (UDP-GlcNAc) is synthesized from glucose via the Hexosamine Biosynthetic Pathway (HBP). Consequently, hyperglycemic conditions metabolically impinge upon the nutrient-sensing HBP [31, 32]. Modification of proteins by O-GlcNAc promotes tumor cell survival, proliferation, and multidrug resistance [33]. In order to determine the impact of a nutrient-dense diet on the overall O-GlcNAc landscape of the mammary tumor, we stained the tumor tissues for O-GlcNAcylation GlcNAcylation (using the RL2 antibody). Tumors from Ob/D mice showed significantly higher levels of RL2 staining compared to tumors from lean mice (p <0.0001). Consistent with the glucose lowering effect, metformin-treated mice showed decreased staining for overall O-GlcNAcylation (p <0.0001) (Figure 4A). The mammary adipose stromal tissue content has been linked to enhanced breast cancer tumorigenicity by affecting the extracellular matrix [34]. In the context of the diet administered, we evaluated tumor sections for their adiposity content. Tumors from HFD-fed mice have significantly greater adipose areas compared to tumors from control diet-fed mice. Interestingly, tumors from metformin-treated mice showed a remarkably reduced adiposity (p=0.014) (Figure 4B). Thus, the overall metabolic status of the host influences mammary tumor O-GlcNAcylation and adiposity. Glycation of collagen alters its physical properties, and promotes tumor cell invasion and metastasis [35]. Tissues of diabetic patients undergo non-enzymatic glycation due to high levels of circulating blood glucose [36]. Furthermore, stromal organization of tumor extracellular matrix influences tumor invasiveness and metastasis [37]. Patients with tumors featuring more aligned collagen tend to show worse survival [37]. We stained tumor tissues for collagen using Picrosirius red and found that tumors from Ob/D mice tend to show a more aligned and organized collagen structure compared to tumors from control lean/non-diabetic mice. Metformin altered this arrangement, resulting in more disoriented, unstructured, and networked collagen fibers (Figure 4C).

Figure 4: Host obesity and diabetes alter tumor O-GlcNAcylation and adiposity.

A. Ob/D tumors showed significantly higher O-GlcNAcylation than the lean tumors (p=<0.0001) Ob/D vs Ob/D+Met (p <0.0001) (one-way ANOVA p <0.0001) O-GlcNAc was evaluated by immunohistochemical staining with RL2 antibody. B. H&E stainingof tumor tissues demonstrates significantly greater amounts of adipose tissue in Ob/D compared to lean and metformin treated mice tumors (p=0.0007, 0.016 respectively). The graph represents quantified percentage of adipose tissues per tumor. C. Tumor sections were assessed for tumor collagen orientation using Picrosirius red staining. Upper panel shows polarized light images, lower panel shows bright field images. Tumors from Ob/D group shows more aligned collagen orientation while lean and Ob/D+Met tumors shows more disoriented networked collagen alignment.

Metformin blunts the malignant features of mammary tumors from diabetic and obese mice

To further investigate how obesity and systemic hyperglycemia impact tumor characteristics, we immunostained these syngeneic mammary tumors for Proliferating Cell Nuclear Antigen (PCNA), a widely accepted marker for proliferation. Ob/D tumors showed a significantly higher population of proliferating cells compared to lean group (p=0.005); interestingly, PCNA staining was significantly reduced in mice treated with metformin (p=0.03) (Figure 5A), indicating that metformin antagonizes proliferation of the rapidly dividing mammary tumor cells. Suppression of apoptosis is an important hallmark of cancer [38] and a desired end point of many targeted therapies to induce tumor cell death [39]. As such, to detect if T2DM alters tumor cell apoptosis, we TUNEL stained these tissues to identify apoptotic DNA fragmentation. Ob/D mice showed significantly fewer tumor apoptotic cells relative to lean mice (p=0.027) (Figure 5B). Notably, Ob/D+Met mice showed a trend towards greater levels of apoptosis compared to Ob/D mice. In addition to these effects on tumor cells, we also examined the effect of Ob/D and metformin on the vascularity of the mammary tumors. To detect vascularization, we stained the tumors with Griffonia Simplicifolia Lectin I, which specifically binds to terminal α-galactosyl residues expressed by mouse endothelial cells. While tumor vascularity was comparable between the tumors from lean mice and Ob/D mice, metformin treated tumors showed significantly decreased vascularity compared to the Ob/D group (p=0.002), indicating a direct effect of metformin on tumor vascularity (Figure 5C). Next, we evaluated the impact of an obese and hyperglycemic host on the metastatic potential of tumors. We examined early metastases in lungs collected from mice 20 days after mammary fat pad injection. Untreated Ob/D mice showed early pulmonary metastatic lesions unlike metformin-treated and non-diabetic mice (Figure 5D), indicating more efficient metastatic seeding. Cumulatively, the data suggest that T2DM supports tumor cell proliferation and simultaneously protects them from apoptosis, fostering metastatic seeding in the lungs. Metformin antagonizes the protective effects of a T2DM on the mammary tumor, and impeded metastasis.

Figure 5: Metformin blunts the malignant features of mammary tumors from diabetic and obese mice.

Tumor sections were assessed for A. Proliferation by PCNA staining, showing significantly lower proliferating cell levels in lean vs Ob/D mice tumors (p=0.005) and in Ob/D+Met vs Ob/D mice tumors (p=0.03). B. Apoptosis by TUNEL staining. Tumors from lean mice had significantly higher levels of apoptotic cells compared to Ob/D group (p=0.027). C. Endothelial cell staining (Griffonia Simplicifolia Lectin I) which shows a significant decrease in vascular density in Ob/D+Met tumors compared to Ob/D tumors (p=0.002). The vascular density was not statistically different between lean and Ob/D tumors. The graph represents quantitated percentage of vascular area per field. D. H&Estaining of mouse lungs showing early metastatic foci in Ob/D groups unlike Lean and Ob/D+Met groups.

Obesity and diabetes reprogram the tumor immune-microenvironment

Gibson et al., recently reported that mammary tumors from obese mice are enriched in granulocytic myeloid-derived suppressor cells [40]. In addition, hyperglycemia has been demonstrated to alter macrophage polarization in muscle tendons [41]. To further characterize the effects of obesity and the hyperglycemic state on the tumor microenvironment, we evaluated the tumor associated macrophage population. Ob/D tumors were characterized by a significantly greater number of CD206 expressing macrophages compared to tumors from lean mice and metformin treated mice (p=0.048 and 0.029, respectively) (Figure 6A). CD206 is a marker of alternatively activated macrophages (M2), which have anti-inflammatory functions that promote tumor proliferation and metastasis [42]. In contrast, the Ob/D tumors showed a significantly lower population of CD80-expressing CD68-positive cells compared to tumors from lean and metformin treated mice (p=0.003, 0.008 respectively) (Figure 6B). CD80 marks classically activated macrophages (M1) which are recognized as inflammatory, anti-tumorigenic immune cells [42]. Accordingly, Ob/D tumors had a significantly higher M2/M1 ratio compared to lean and metformin treated mice (p=0.03 both) (Figure 6C). Tumors bearing a high M2/M1 ratio tend to show a more aggressive phenotype and enable a tumor aiding immune microenvironment [43]. Thus, while obesity and diabetes impinge upon a tumor-permissive macrophage population, metformin was able to reverse this trend, providing evidence for metformin in also impacting the tumor immune microenvironment.

Figure 6: Obesity and diabetes reprogram the tumor immune-microenvironment.

A. Tumor sections from lean, Ob/D, and Ob/D+Met mice were assessed for M2 macrophages by immunofluorescence staining for CD206 (M2 polarized macrophages). Tumors from Ob/D mice show a significant increase (p=0.02) in the abundance of CD206+ cells. Metformin treatment decreases this significantly (p=0.02). The graph represents quantification of CD206+ cells/field. B. Tumor sections from lean, Ob/D, and Ob/D+Met mice were assessed for M1 macrophages by dual immunohistochemistry for CD80/68 staining. CD68 marks all macrophages, while CD80 depicts M1-polarized macrophages. Tumors from lean mice show significantly greater numbers of infiltrating CD80+/CD68+ cells relative to Ob/D tumors (p=0.003). Metformin-treated tumors show a significant increase in M1 macrophages relative to Ob/D mice tumors (p=0.008). The graph represents quantification of double positive cells/field. C. The graph represents a ratio of quantified M2/M1 polarized macrophages per field in tumor sections. The results are statistically significant (One-way ANOVA, p=0.01)

Discussion

Women with T2DM are often diagnosed with advanced stage breast cancer [9, 44–48]. Hyperglycemia is one of the main characteristics of diabetes and is considered a possible reason for worsened prognosis and increased risk of death of breast cancer patients [49, 50]. Tumors from women with invasive ductal breast carcinoma who also present with T2DM, are larger and are accompanied by more frequent lymphatic and distant metastases compared to those without T2DM (p<0.05). These tumors also express significantly elevated levels of matrix metalloproteinases, MMP2 and MMP9, that enable their invasive, metastatic spread [51]. Among all the breast cancer subtypes, diabetic patients are at a higher risk of triple-negative breast cancer (TNBC) relative to non-diabetic patients [50, 52–56]. Around 35–40% of TNBC patients present with this co-morbid diabetic condition [57]. In TNBC patients, T2DM is an independent prognostic risk factor that indicates an increased likelihood of recurrence and metastasis [52, 53].

The complications of T2DM are further confounded by co-presentation of obesity [58]. Together, obesity and diabetes are associated with poor prognosis of breast cancer [59].

These studies agree with our analytic cohort of female BC patients, where 17.5% of patients had a diagnosis of T2DM and 53.3% were obese or severely obese (BMI≥30), giving an insight about the prevalence of these complications in BC patients in the state of Alabama. The survival probability was significantly greater in non-diabetic versus diabetic breast cancer patients (93.2% vs 85.5%) and in non-obese versus obese patients (92.6% vs 90.4%). Overall, these findings underscore the significance of characterizing and understanding the nature of tumors that grow in the context of these distinct metabolic conditions in the host.

Cancer cells usually have higher rates of glucose uptake to serve as a fuel for tumorigenesis and growth [60] This might explain why a state of chronic hyperglycemia is associated with poorer prognosis in cancer patients [61]. Other mechanisms that might be contributing to this observation as well are hyperinsulinemia, oxidative stress and deviant sex hormone production [61]. Mechanistically the tumor promoting effects of hyperglycemia manifest as enhanced tumor cell proliferation, invasion, migration, and apoptosis resistance [61].

We employed the C57BL/6J mouse strain since it is readily amenable to diet-induced obesity and insulin resistance [62] [27]. These mice develop obesity, hyperinsulinemia, hyperglycemia, and hypertension when allowed ad libitum access to a diet with high fat content [27]. Utilizing this mouse strain, we developed a syngeneic mouse model of diet-induced obesity and hyperglycemia that simulates type 2 diabetes [23] with triple-negative mouse mammary carcinoma and investigated the effect of metformin on tumor characteristics and the microenvironment in relation to its glucose lowering properties. Ob/D mice supported larger tumors that reflected the metabolic status of the host, demonstrating remarkable intra-tumoral adipose tissue compared to mice fed a control diet or treated with metformin. This is suggestive of the possibility that metformin not only contributes to controlling hyperglycemia but may also reduce tumor adiposity. Metformin is effective in reducing body weight in insulin sensitive and insulin resistant overweight and obese patients [63, 64]. It is likely that in addition to its systemic effects, metformin reconfigures an overall anti-tumor metabolic environment in obese cancer patients. Metabolic perturbations including hyperglycemia and diabetes promote O-GlcNAc modifications on key signaling molecules and transcription factors associated with diabetic conditions, neurodegenerative disease, and cancer [65–68].Elevated glucose is shunted through the HBP and culminates in increased O-GlcNAcylation ofproteins. Tumors from obese and diabetic untreated mice showed a metabolic switch characterized by higher O-GlcNAc levels compared to tumors in mice fed a control diet; this reflects the role of HBP in relaying the state of host nutrient abundance to the tumor cells, contributing to the mechanism by which hyperglycemia promotes tumor progression and metastasis. Metformin mediated AMPK activation tempers cellular O-GlcNAcylation [12, 13]. In fact, O-GlcNAc inhibition is the main mechanism by which AMPK blocks cardiac hypertrophy [12]. In our study, in obese and diabetic mice, metformin caused a decline in the O-GlcNAc landscape of tumors, further evidencing metformin’s multimodal activities in impacting multiple signaling pathways. In contrast to its role in regulating O-GlcNAcylation, there are conflicting reports regarding metformin’s role on endothelial cells and angiogenesis. While metformin attenuates pro-angiogenic and inflammatory stimuli such as tumor necrosis factor α (TNFα), nuclear factor-κB (NF-κB), plasminogen activator inhibitor-1 (PAI-1) antigen, and von Willebrand factor [69–71], its role in cardiovascular disease models is associated with vascular endothelial growth factor (VEGF) upregulation and an increase in nitric oxide bioavailability, both of which support enhanced angiogenesis [72–74]. Functionally, metformin inhibits formation of capillary-like networks by endothelial cells; this effect is partially dependent on the AMPK energy sensor [75]. Moreover, metformin’s effects are cell context-dependent – acting differently in tumor cells and endothelial cells, likely because the activity of AMPK is differently regulated in tumor and normal cells [75]. While tumors from our obese and diabetic mice were richly vascularized, metformin-treated tumors showed a significantly lower vascular density. This effect on tumor vasculature is likely due to a direct effect of metformin on tumor cells. It is possible that metformin inhibits precursor endothelial cells (neoangiogenesis) or interferes with the angiogenesis process in the context of obesity and/or diabetes.

By way of suppressing excessive deposition of extracellular matrix in adipose tissue, metformin ameliorates insulin resistance in obesity [76]. It also is likely that metformin may mitigate glycation of the collagen matrices in a state of diabetic hyperglycemia [35]. In the context of obesity and diabetes, metformin altered the collagen architecture of the mammary tumor. This is an important consideration because structural or biophysical properties of the extracellular matrix can alter the fate of the tumor. Collagen I fibril alignment influences tumor cell morphology, invasiveness, and cluster formation. In particular, while disoriented collagen fibers present a physical hurdle, aligned or bundled collagen fibrils promote tumor cell invasiveness and migration leading to metastasis. While in vitro studies demonstrate that hyperglycemia promotes tumor cell invasiveness [77], the architecture of the tumor matrix presents a tumor cell-autonomous variable that can either facilitate or impede invasion, intravasation, and metastasis [78]. Metformin-treated tumors presented with a networked collagen morphology, suggesting a broader role for metformin in restructuring the flexible tumor matrix.

Structural, compositional, and metabolic alterations in the tumor microenvironment contribute to an intense reprogramming process that modulates immune cells [79]. The tumor immune microenvironment encompasses complex tumor suppressing and tumor promoting functions, which eventually affect tumor growth and metastasis [80]. Glucose metabolism particularly has been shown to modulate macrophage activation and polarization [81]. Tumor associated macrophages are involved in diverse mechanisms including wound healing and inflammation; for this reason macrophages are able to implement different activation states based on the stimulus [82]. Monocyte derived macrophages are polarized into either classically activated macrophages (M1), which are inflammatory in nature, and provide resistance to tumor progression or into alternatively activated macrophages (M2), which are anti-inflammatory and aid angiogenesis to provide a tumor promoting environment [83]. Breast tumors with high, M2,-like macrophages or high M2/M1 ratios tend to be more aggressive and foster a tumor aiding immune microenvironment [43] [42]. Tumors from obese and diabetic mice had a greater infiltration by CD206-expressing M2-polarized macrophages simultaneous with a smaller population of inflammatory CD80/68 expressing M1-polarized macrophages. Interestingly, metformin reversed this trend. Our findings are in contrast with literature supporting a therapeutic role for metformin in inducing M2 polarization to promote wound healing and reduce systemic inflammation [84, 85]. However, in the context of glioblastoma, metformin promoted M1 polarization and prolonged the pro-inflammatory effects of both low dose and high dose radiation [86]. In a melanoma model, metformin increased the abundance of activated CD8+ T cells (expressing Granzyme B) and the numbers of CD3+ effector-memory CD8+ T cells [87]. Thus, metformin appears to present with paradoxical effects both systemically, and in the tumor. In agreement with decreased tumor vascularization, collagen remodeling, and an abundance of tumor infiltrating anti-inflammatory macrophages, we saw that metformin decreased the incidence of early metastasis in obese and diabetic mice. While hyperglycemia has been shown to impair neutrophil mobilization favoring metastatic seeding [88], our data presents a role for macrophages in creating a tumor-permissive environment for growth and metastatic dissemination.

Collectively, our data show that underlying obesity and hyperglycemia facilitate primary tumor growth and a remarkable re-structuring of the tumor microenvironment, concomitant with an increase in early metastatic seeding. Cumulatively, the underlying metabolic conditions foment an aggressive tumor phenotype. We show that normalizing glucose levels reconfigures the mammary tumor milieu and decreases metastasis. In the era of precision medicine, that seeks to develop personalized treatments according to a patient’s tumor characteristics, more understanding is needed to characterize breast cancer tumors that grow in the exceptional diabetic and obese metabolic background. Given the anticipated public health problem of rising prevalence of diabetes and obesity in our population, we may very well be anticipating a concurrent upsurge of aggressive and metastatic breast cancers.

Abbreviations

T2DM

Type 2 diabetes mellitus

AUC

Area under the curve

BLG

Blood Glucose

BLI

Bioluminescence imaging

ECM

Extracellular matrix

GTT

Glucose tolerance test

HBP

Hexosamine biosynthetic pathway

HFD

High fat diet

IF

Immunofluorescence

IHC

Immunohistochemistry

IP

Intraperitoneal

M1

Classically activated macrophages

M2

Alternatively activated macrophages

Ob/D+Met

Metformin treated diabetic obese mice

MFP

Mammary fat pad injection

Ob/D

Obese diabetic

O-GlcNAc

O-linked β-N-acetylglucosamine

PCNA

Proliferating Cell Nuclear Antigen

STZ

Streptozotocin

TNBC

Triple-negative breast cancer

VEGF

Vascular endothelial growth factor

WHO

World Health Organization