Metformin inhibits RANKL and sensitizes cancer stem cells to denosumab

ABSTRACT

The increased propensity of BRCA1 mutation carriers to develop aggressive breast tumors with stem-like properties begins to be understood in terms of osteoprotegerin (OPG)-unrestricted cross-talk between RANKL-overproducing progesterone-sensor cells and cancer-initiating RANK⁺ responder cells that reside within pre-malignant BRCA1^mut/+ breast epithelial tissue. We recently proposed that, in the absence of hormone influence, cancer-initiating cells might remain responsive to RANKL stimulation, and hence to the therapeutic effects of the anti-RANKL antibody denosumab because genomic instability induced by BRCA1 haploinsufficiency might suffice to cell-autonomously hyperactivate RANKL gene expression. Here we report that the biguanide metformin prevents BRCA1 haploinsufficiency-driven RANKL gene overexpression, thereby disrupting an auto-regulatory feedback control of RANKL-addicted cancer stem cell-like states within BRCA1^mut/− cell populations. Moreover, metformin treatment elicits a synergistic decline in the breast cancer-initiating cell population and its self-renewal capacity in BRCA1-mutated basal-like breast cancer cells with bone metastasis-initiation capacity that exhibit primary resistance to denosumab in mammosphere assays. The specific targeting of RANKL/RANK signaling with denosumab is expected to revolutionize prevention and treatment strategies currently available for BRCA1 mutation carriers. Our findings provide a rationale for new denosumab/metformin combinatorial strategies to clinically manage RANKL-related breast oncogenesis and metastatic progression.

KEYWORDS:

• BRCA1
• breast cancer
• denosumab
• metformin
• RANK
• RANKL

This article refers to:

The cytokines receptor activator of nuclear factor κB ligand (RANKL) and osteoprotegerin (OPG) play key roles in the differentiation and function of osteoblasts, cells that are specialized in the degradation of bone tissue.^1-3 Binding of RANKL to its receptor, RANK, on the cell surface of osteoclasts or their precursors induces differentiation, activation, and survival, and stimulates bone resorption. OPG is the endogenous decoy receptor for RANKL that antagonizes RANKL/RANK-mediated signaling and hence inhibits bone resorption, preserving bone density. The balance between RANKL and OPG is important for bone metabolism and its dysregulation is a major driving force in cancer treatment-induced bone loss (CTIBL) and in the development of malignant bone lesions.⁴ Not surprisingly, the RANKL/OPG signaling axis has emerged as an attractive target not only to prevent bone loss, but also to prevent bone metastases in breast cancer.^5,6

Denosumab, a fully human antibody targeting RANKL, is a paradigmatic example of the family of bone resorption inhibitors developed to treat osteoporosis and to prevent skeletal damage caused by bone metastases. Bisphosphonates such as zoledronic acid are the current standard of care for the treatment of skeletal-related events (SREs), including bone metastases, in patients with advanced breast cancer.^3,4 However, results from clinical studies suggest that denosumab efficacy, as determined by a reduction in SREs, is at least equal to bisphosphonates and possibly superior in a large prospectively randomized phase III trial of advanced breast cancer patients.^3-5 Ongoing clinical studies evaluating the effects of denosumab in CTIBL and on bone metastasis-free survival in women with early-stage breast cancer at high risk for disease recurrence should confirm whether it can challenge the bisphosphonate-based current standard care of bone metastasis in breast cancer patients. The repurposing of an osteoclastogenesis inhibitor such as denosumab may herald a new era of primary prevention therapies, and perhaps cancer treatments, in woman carriers of BRCA1 gene mutations.

Dysregulation of the OPG/RANKL/RANK signaling axis appears to play a role in breast carcinogenesis that may be particularly relevant for BRCA1-mutation carriers facing a high lifetime risk of breast cancer.^7-10 The increased propensity of this population to develop aggressive breast tumors with stem-like properties begins to be understood in terms of OPG-unrestricted cross-talk between RANKL-overproducing cells and cancer-initiating RANK⁺ cells that reside within pre-malignant breast epithelial tissues.^11,12 Free serum levels of OPG are significantly lower in BRCA1 mutation carriers than in non-carrier controls. Moreover, among BRCA1 carriers, lower OPG levels are associated with germline mutations known to confer an increased risk for breast cancer. High plasma OPG levels in BRCA1-mutation carriers correlate with a significantly decreased risk of developing breast cancer as compared with BRCA1-deficient women with low OPG levels.^7-9 These findings point to a promising role for OPG as a biomarker to identify BRCA1 mutation carriers who are at higher risk of developing breast cancer and who would benefit from therapies mimicking the physiological role of OPG, such as the anti-RANKL antibody denosumab.¹⁰

The net magnitude of the OPG/RANKL/RANK signaling axis in driving BRCA1-related breast carcinogenesis seems to be determined by the decrease in local and systemic OPG along with the local increase of RANKL, thus supporting the relevance of therapeutically manipulating OPG levels in BRCA1 carriers by directly targeting RANKL. Along this line, 2 highly publicized studies have recently reported that RANKL inhibition, achieved by using either a recombinant OPG-Fc fusion protein or RANKL-specific monoclonal antibodies such as denosumab, can significantly delay tumor onset in clinically relevant models of BRCA1 deficiency.^{11, 12} Mechanistically, denosumab is thought to interfere with the cross-talk between RANKL-producing “sensor” cells and cancer-initiating RANK⁺ responder cells that appear to reside within pre-malignant tissues of BRCA1-mutation carriers. We recently provided support to the alternative but not mutually exclusive hypothesis that BRCA1 haploinsufficiency may be sufficient to cell-autonomously activate RANKL expression and generate breast cancer-initiating cell populations responsive to denosumab.¹³

Metformin prevents BRCA1 haploinsufficiency-driven upregulation of RANKL in breast epithelial cells. The anti-diabetic biguanide metformin is consistently associated with a decrease in cancer incidence and mortality, both overall and also linked to specific organ sites including breast.^14-16 Metformin has also been shown to exert direct osteogenic effects by enhancing osteoblast differentiation and inhibiting osteoclast differentiation in vitro and by preventing bone loss in ovariectomized rats.^17-21 The anti-osteoclastogenic activity of metformin seems to involve the stimulation of OPG production and the reduction of RANKL expression by osteoblasts. We recently provided unbiased evidence that mono-allelic inactivation of BRCA1 in normal-like breast epithelial cells suffices to cell-autonomously generate RANKL-overproducing, denosumab-responsive cancer stem cell (CSC)-like cellular states.¹³ Thus, we envisioned that the unanticipated capacity of metformin to regulate breast epithelia-associated RANKL expression might be a new approach to target RANKL/RANK signaling that generates and maintains CSC-like properties in BRCA1-deficient breast cancer cells.

We recently reported activation of RANKL gene expression in normal-like MCF10A human breast epithelial cells containing the 185delAG mutation in one BRCA1 allele.¹³ This cell line is a good model of genomic instability and recapitulates the cell-autonomous consequences of one-hit BRCA1 inactivation in breast epithelia of carriers of BRCA1mutations.^22-24 We here assessed the impact of metformin treatment on the expression status of RANK and RANKL in BRCA1^+/+ parental cells and in BRCA1^185delAG/+ haploinsufficient isogenic derivatives. RT-PCR analysis showed that metformin treatment (5 mmol/L) moderately reduced the expression of RANK in both cell lines (Fig. 1). Remarkably, the ability of metformin to substantially reduce RANKL gene expression in BRCA1^+/+ parental cells (4-fold reduction vs. untreated BRCA1^+/+ cells) remained unaltered in a background of BRCA1 deficiency. Accordingly, the very high levels of RANKL gene expression triggered by BRCA1 haploinsufficiency were completely suppressed (5-fold reduction vs. untreated BRCA1^185delAG/+ cells) by metformin (Fig. 1). These findings reveal for the first time that metformin might target the hyperactivation of the RANKL/RANK signaling axis by preventing BRCA1 deficiency-induced cell-autonomous hyperactivation of RANKL gene expression.

Figure 1. Metformin inhibits RANKL gene expression in breast epithelial cells. Total RNA from untreated and metformin-treated (5 mmol/L; 48 h) BRCA1^+/+ and BRCA1^mut/+ MCF10A isogenic cell pairs was evaluated in technical triplicates for the abundance of RANK (TNFRSF11A, Hs00921372_m1) and RANKL (TNFSF11, Hs00243522_m1) relative to housekeeping genes GADPH (Hs99999905_m1) and 18S (Hs99999901_s1). The transcript abundance was calculated using the delta Ct method and presented as relative quantification (RQ) or log2 fold-change, as specified. MET, Metformin.

Metformin synergistically sensitizes BRCA1-deficient breast cancer stem cells to denosumab. Our previous findings showing the robust ability of metformin to reduce the mammosphere-initiating capacity of BRCA1^185delAG/+ cells did not include an analysis of the nature of the interaction between metformin and denosumab.^{13, 24} Thus, we sought to explore the therapeutic relevance of metformin for the RANKL-dependent ability of denosumab to reduce tumor-initiating capacity of triple-negative MDA-MB-436 cells. Metastatic MDA-MB-436 cells contain a BRCA1 5396 + 1G>A mutation, resulting in an in-frame deletion of 28 amino acids and a truncated BRCA1 protein,²⁵ and they are capable of developing osteolytic mammary tumors upon direct intra-tibial injection into immunocompromised mice.^{26, 27} Since the number and the size of mammospheres reflects the quantity of CSC-like cells capable of self-renewal and the self-renewal capacity of each CSC-like cell, respectively, we used the recently developed Cell2Sphere™ Kit (StemTek Therapeutics, Bilbao, Spain) to evaluate the impact of denosumab and metformin on the number and size of mammospheres formed by BRCA1-mutant MDA-MB-436 cells. Treatment with either a clinically relevant concentration of denosumab [1 μg/mL; 28] or metformin (1 mmol/L) as single agents moderately reduced (∼20% and 10%, respectively) mammosphere-forming efficiency (MSFE) of MDA-MB-436 cells (Fig. 2). Remarkably, the concurrent combination of denosumab and metformin supra-additively decreased MSFE by greater than 50%. Moreover, under these conditions mammospheres were markedly smaller than those generated from untreated control cells; indeed, there were no mammospheres >100 μm in size (Fig. 2).

Figure 2. Metformin sensitizes mammosphere-initiating cells to denosumab. Cell2Sphere™ assays using BRCA1-deficient MDA-MB-436 cells were performed as per the manufacturer's instructions ((http://stemtektherapeutics.com/en/cell2sphere#cell2sphere_kit). Drugs were added to quintuplicate sets of wells on days 1 and 4 without replenishing the medium. ImageJ was used to quantify the size (left; central lines indicate mean values) and number (right) of 6-day-old mammospheres. *P < 0.05, **P < 0.001 MET, Metformin. Size bar = 2000 μm.

The statistically significant reduction in the number and size of mammospheres generated with denosumab plus metformin but not with denosumab alone indicates that metformin functions synergistically to decrease the magnitude of the mammosphere-initiating population as well as its self-renewal capacity in denosumab-refractory CSC-like cells.

Metformin-regulated RANKL: From cancer prevention to circumventing denosumab resistance. Beyond prophylactic mastectomy, there are currently few options available to reduce the risk of developing breast cancer for BRCA1 mutation carriers.²³ We are rapidly accumulating evidence that susceptibility to oncogenesis and tissue specificity in BRCA1 mutation carriers is, at least in part, a net consequence of cell nonautonomous and cell-autonomous factors of the OPG/RANKL/RANK axis.

Whereas aberrant DNA damage-triggered hyperactivation of the RANKL/NF-κB signaling axis might account for the expansion of cell populations within BRCA1^mut/+ pre-neoplastic breast tissue with breast cancer-initiating capability, low systemic levels of the natural antagonist OPG would provoke an unrestrained amplification of such genetically unstable populations of BRCA1 haploinsufficient cells with ensuing DNA damage activation of RANKL/NF-κB signaling. Such a model of epithelium hyperproliferation during breast oncogenesis in pre-menopausal BRCA1 mutation carriers likely involves a progesterone-inducible secretion of RANKL in so-called sensor cells (i.e., mature ductal cells) and subsequent RANKL stimulation of so-called “responder” breast cancer-initiating RANK+ cells (i.e., hormone receptor-negative luminal progenitors).²⁹ Our recent finding that triple-negative/basal-like breast epithelial cells, engineered to have impaired DNA repair and consequent genomic instability upon targeted mutational knockout of BRCA1 single allele, overproduce RANKL strongly suggests that RANKL-addicted cell populations with breast cancer-initiating capacity would persist during hyperplasia, even in the absence of hormonal influence and despite the drop in RANKL expression occurring in the post-menopausal setting.¹² An idoneous approach for breast cancer prevention might therefore involve strategies to revert the chronically aberrant high RANKL:OPG ratio in breast epithelium of BRCA1-mutation carriers by increasing systemic (serum) OPG while decreasing local RANKL in epithelial cells (Fig. 3).

Figure 3. Dysregulation of the OPG/RANKL/RANK signaling axis and BRCA1 deficiency-driven breast oncogenesis: (A)new target for anti-cancer metformin. Molecular dialog between progesterone (Pg)-responsive cells and hormone receptor (HR)-negative luminal progenitors seems to drive the hyperproliferation of breast cancer-initiating cell populations within the breast epithelia of BRCA1 mutation carriers.^11,12 In a hormone-dependent stage (e.g., pre-menopausal women), RANKL secretion might become stimulated upon binding of Pg to its receptor (PR) in mature ductal cells (i.e., so-called “sensor” cells); secreted RANKL then binds and activates RANK on RANK+ luminal progenitors (i.e., so-called “responder” cells). Generally, OPG negatively regulates RANKL/RANK signaling by binding to RANKL as an endogenous decoy receptor, thereby inhibiting signal transduction. However, the mitogenic cross-talk between RANKL-producing sensor cells and RANKL-addicted responder cells might be less restricted in high-risk subgroups of BRCA1 mutation carriers with significant lower plasma levels of OPG. The net magnitude of the OPG/RANKL/RANK signaling axis in driving BRCA1-related breast oncogenesis may be largely determined by local increases of RANKL along with decreases in systemic OPG. Our recent findings in HR-negative, basal-like MCF10A breast epithelial cells with a pathogenic185delAG mutation in one BRCA1 allele¹³ further suggest that, in the absence of hormone influence, breast cancer-initiating cells may remain responsive to RANKL stimulation because compromised DNA repair and consequent genomic instability provoked by BRCA1 haploinsufficiency might suffice to establish hyperplasic states via an auto-regulatory feedback loop triggered by cell-autonomous hyperactivation of RANKL gene expression. Such a hormone-independent scenario might become reinforced by low circulating levels of OPG not only in post-menopausal BRCA1 carriers but also in other post-menopausal groups at high risk for breast cancer and in women with circulating tumor cells. The ability of metformin (MET) to restore dysregulated OPG/RANKL/RANK signaling may represent an important clinical opportunity to prevent breast cancer and maintain bone health (e.g., following salpingo-oophorectomy) in BRCA mutation carriers and to treat RANKL-driven metastasis-initiating lesions in combination with denosumab.

Preclinical models of inherited cancer syndromes are increasingly appreciated as valuable experimental systems to develop chemopreventive agents (e.g., denosumab and metformin) for application in hereditary and sporadic cancer settings.³⁰ While it is acknowledged that the potential of denosumab as a non-surgical preventive therapy to delay or prevent BRCA1-related breast cancer should await the results of large randomized clinical trials, the ability of metformin to potentially lower the RANKL/OPG ratio, by inhibiting local RANKL gene expression and/or increasing systemic levels of OPG, warrants further exploration as a combinatorial strategy that might prove safe and effective for pre-menopausal BRCA1 mutation carriers before they consider mastectomy. Such a combination could potentially have broader applicability, including post-menopausal BRCA1 carriers and others at increased genetic or familial risk. Moreover, because metformin has been shown to significantly increase total body bone mineral density and prevent bone loss in ovariectomized adult rats,¹⁹ its ability to affect bone metabolism³¹ together with the RANKL-targeting activities of metformin and denosumab might synergize to minimize the loss in bone mineral density that accompanies ovarian cancer-preventive bilateral salpingo-oophorectomy in BRCA mutation carriers.

The ability of denosumab to reduce mammosphere forming capacity--an accepted surrogate for the relative proportion and survival of CSCs--not only in BRCA1 haploinsufficient breast epithelial cells but also in triple-negative/basal-like and HER2-overexpressing breast cancer cells,¹³ supports the notion that tumor- and metastasis-initiating cells from genetically diverse breast cancer subtypes might share a common addiction to hyperactive RANKL signaling in primary and metastatic locations including the bone microenvironment.³² Beyond prevention, the exploration of a role for therapeutic targeting of RANKL with denosumab for the treatment of established breast cancer might take advantage of the expected incorporation of new biomarkers capable of stratifying breast cancer patients that will benefit from bone-targeting, metastasis-preventing treatments such as denosumab from those that, because of de novo or acquired mechanisms of escape, require alternative or additional sensitizing therapies.³³ Triple-negative/basal-like MDA-MB-436 cells with well-recognized bone metastasis-initiation capacity^{26, 27} were found to behave as denosumab-resistant breast cancer cells in terms of tumorsphere-initiating capacity. Remarkably, they became sensitized to the RANKL-targeting activity of denosumab in the presence of metformin. Intriguingly, denosumab-resistant MDA-MB-436 cells exhibit a high copy number of the proto-oncogene c-MAF,³⁴ a recently discovered bone metastasis biomarker that might help to distinguish responders and non-responders to RANKL-targeted therapies.³⁵

The specific targeting of autocrine/paracrine RANKL/RANK signaling with denosumab is expected to revolutionize prevention and treatment strategies currently available for BRCA1 mutation carriers. We propose that the ability of metformin to: a) suppress breast epithelial RANKL gene expression; b) impair the breast cancer-initiating capacity of multiple breast cancer subtypes including those of cancer-prone BRCA1^mut/+ haploinsufficient stages,^{24, 36-38} and c.) synergize with the RANKL-dependent actions of denosumab in breast cancer cells with bone metastasis-initiation abilities, may provide a strong rationale for new denosumab/metformin combinatorial strategies to clinically manage RANKL-related breast oncogenesis and metastatic progression.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgment

The authors would like to thank Dr. Kenneth McCreath for editorial support.

Funding

This work was supported by grants from the Ministerio de Ciencia e Innovación (Grant SAF2016–80639-P to J. A. Menendez), Plan Nacional de I+D+I, Spain and the Agència de Gestió d'Ajuts Universitaris i de Recerca (AGAUR) (Grant 2014 SGR229 to J. A. Menendez), Departament d'Economia i Coneixement, Catalonia, Spain. Elisabet Cuyàs is supported by the Sara Borrell post-doctoral contract (CD15/00033) from the Ministerio de Sanidad y Consumo, Fondo de Investigación Sanitaria (FIS), Spain. Joaquim Bosch-Barrera is supported by an Emerging Research Grant (2013) from the Spanish Society of Medical Oncology (SEOM, Madrid, Spain), and a Research Grant from Pfizer (WI190764).