Zoledronic acid effectiveness against breast cancer metastases - a role for estrogen in the microenvironment?

Zoledronic acid (ZA) is an imidazole-containing bisphosphonate that has been extensively studied as an osteoclast inhibitor. ZA decreases bone turnover and has been effective in limiting osteolysis in metastatic cancers, including breast cancer. Recent clinical trials that demonstrated enhancement of disease-free survival by bisphosphonates have prompted interest in bisphosphonates as anti-cancer agents. ZA, for example, increased disease-free survival in postmenopausal and in premenopausal, hormone-suppressed breast cancer patients. Intriguingly, however, there was a lack of an anti-cancer effect of ZA in premenopausal women without ovarian suppression. These observations have prompted the conjecture that anti-cancer effects of ZA are limited to estrogen-poor environments. This review explores possible mechanisms compatible with differences in ZA activity in premenopausal women compared with postmenopausal (or hormone-suppressed) women.


In the Austrian Breast & Colorectal Cancer Study
Group 12 (ABCSG12) study, 3 years of ZA added to endocrine therapy increased the DFS of premenopausal women with stage I/II breast cancer compared with endo crine therapy alone. All of the premenopausal women in the ABCSG12 trial received luteinizing hormone-releasing hormone agonist goserelin to suppress production of estrogen and progesterone. Production or function of estrogen was further inhibited by anastrozole or by tamoxifen. Th e study demonstrated a 36% reduction overall in the relative risk of disease progression among those patients taking ZA. Th e increased DFS was sustained after treatment cessation to 82 months [4]. Notably, the ABCSG12 trial demonstrated that ZA reduced both intra-osseous and extra-osseous breast cancer recurrence and locoregional recurrences in these premenopausal women on hormone suppression.
In contrast, the Adjuvant Zoledronic Acid to Reduce Recurrence (AZURE) trial indicated that addition of ZA to standard therapy did not prolong DFS in women with stage II/III breast cancer [5]. Th e AZURE study enrolled 2,259 premenopausal or perimenopausal women and 1,101 postmenopausal women, nearly all of whom received chemotherapy, and randomized them to adjuvant ZA treatment or placebo. ZA did not improve the overall survival or DFS of the entire cohort, but a prespecifi ed subgroup analysis in the AZURE study indicated that postmenopausal women -but not premenopausal women -treated with ZA showed an increased DFS of a magnitude comparable with that observed in the ABGCS12 trial [5]. ZA was eff ective in preventing new secondary primary tumors and locoregional and nonskeletal distant recurrences in postmenopausal women in the AZURE trial.
Although not the primary endpoint, a DFS benefi t of ZA treatment in postmenopausal women has also been supported by trials in which postmenopausal women with early-stage hormone-responsive breast cancer were treated with the aromatase inhibitor letrozole and were randomized to either immediate or delayed ZA (the ZO-Fast, Z-Fast, E-ZO-Fast and N03CC trials). Delayed ZA therapy was triggered by nontraumatic fracture or crossing a bone loss threshold. Th e results not only demonstrated improved bone mineral density (the primary endpoint) with immediate ZA treatment, but showed a benefi t for DFS -in the ZO-Fast trial at 48 months [6] and in the Z-Fast trial at 12 to 48 months but not at 5 years [7] (the two other similar trials (E-ZO-Fast and N03CC) had insuffi cient data to evaluate recurrence at 12 months [8,9]). Additional information is presented in Table 1.
Collectively, these trials have led to several conclusions and potential interpretations. First, there is evidence of an anti-cancer eff ect of ZA that is not limited to bone. Second, ZA opposes breast cancer recurrence in sexhormone-poor environments (that is, goserelin plus antiestrogens or in postmenopausal women). Although the ABCSG12 and AZURE trials diff er in enrollment criteria and the extent of adjuvant chemotherapy usage, the benefi t of ZA in premenopausal women in the ABCSG12 trial and not in the AZURE trial has been interpreted as an inhibitory eff ect of sex hormones on the ability of ZA to augment DFS [5]. Th ird, the lack of anti-cancer eff ectiveness of ZA in premenopausal women could not be rescued by tamoxifen alone, given that 74% of subjects in each arm of the AZURE trial received tamoxifen plus chemotherapy [5] (R Coleman, personal communication). Opposition of ZA improvement of DFS could therefore be mediated by tamoxifen-insensitive estrogen receptor (ER) activity or through the activity of other, goserelinsensitive, hormones. Finally, results from the AZURE trial showed that premenopausal women responded poorly to ZA -independent of the ER status of the tumor. ZA resistance in premenopausal women may there fore not arise from estrogen or progesterone signal ing within the cancer cells themselves, but rather from sex hormone eff ects on the premenopausal micro environment.
In this review we consider how these interpretations could be informed by preclinical in vitro and in vivo studies of ZA activity, focusing on the possible endocrinedependent eff ects of ZA activity. In the AZURE trial, premenopausal women with either ER-positive or ERnegative tumors responded less well to ZA than did postmenopausal women [5]. Th is fi nding indicates that sex hormones did not solely antagonize ZA by acting on cancer cells but through hormonal manipulation of stroma or the tumor microenvironment, which we will discuss here. Although a number of hormone receptor pathways could be involved in the diff erential ZA responses, we will focus on estrogen, and only briefl y discuss other hormones such as progesterone and inhibin.

Eff ects of zoledronic acid in bone
ZA, mimicking pyrophosphate, binds to hydroxyapatite in mineralized bone, and is directly toxic to osteoclastsit limits bone turnover and skeletal-related events arising from bone-metastatic cancer. By inhibiting osteoclastmediated bone resorption, ZA lowers bone marrow calcium and impedes the liberation of growth factors from the bone matrix that contribute to the vicious cycle of metastasis growth and bone breakdown. Th ere are some data suggesting that bisphosphonates can have additional, osteoclast-independent eff ects. For example, ZA was shown to be eff ective against bone tumors in mice with nonfunctional osteoclasts [10]. However, ZA's primary eff ect on osteoclasts is supported by data in patients with advanced breast cancer showing that ZA-treated breast cancer patients in whom ZA rapidly normalized bone turnover (manifested by N-telopeptide of type I collagen) had longer event-free survival and overall survival than those with continued high turnover [11].
In addition to the inhibition of osteoclast activity, ZA was shown to have many other activities within bone itself that could explain its ability to prevent recurrence of breast cancer. ZA downregulates adhesion molecules on bone marrow stromal cells and inhibits tumor cell adhesion to the bone matrix whether the tumor cells or the matrix are treated with bisphosphonate [12]. ZA also promotes osteoblast maturation and production of the rank ligand antagonist osteoprotegerin, an inhibitor of osteolysis [13]. Moreover, ZA has been shown to inhibit stromal IL-6 secretion and matrix metallo proteinase-1 production, eff ects that could block tumor growth in bone or at other sites.
Th ere is increasing evidence that ZA could aff ect survival of disseminated tumor cells, whose prevalence predicts local and distant relapse [14]. Both neoadjuvant and post chemotherapy administration of ZA (and other bisphos phonates) have been shown to decrease disseminated tumor cells in marrow compared with pretreatment levels [15,16], lowering the risk of metastatic spread from micrometastases. Th is occurrence could represent an anti-cancer eff ect. Interestingly, osteoclasts and osteoblasts have been implicated in the maintenance of dormant leukemic clones in the marrow [17], which may also be the case for breast cancer. Moreover, breast tumor cell viability in the endosteal niche has been shown to involve heterotypic notch/ jagged interactions with osteoblasts similar to those used to regulate hematopoietic stem cell numbers [18,19].
ZA-mediated shifts in osteoclast and osteoblast function may therefore lead to changes in the microenvironment that undermine support for disseminated cancer cells.

Estrogen actions in bone as a modulator of zoledronic acid eff ectiveness
As described above, the clinical data point towards a role for low-estrogen environments in high ZA effi cacy. A possible caveat is the clinical observation that most of the premenopausal women in the AZURE trial who reaped no benefi t from ZA were receiving tamoxifen. Conceivably, tamoxifen did not adequately block estrogen signaling that undermined ZA (discussed below); alternatively, the critical action of estrogen impairing the ZA anti-tumor eff ect occurred at a site where tamoxifen functioned as an agonist (for example, bone). Like estrogen, tamoxifen opposes bone turnover (albeit to a lesser degree [20]), and thus suppression of bone turnover by tamoxifen or estrogen may limit the magnitude of benefi t accruing from ZA treatment because ZA has a greater survival benefi t in patients with bone metastases with high basal levels of bone turnover [21].
Studies have not yet uncovered a mechanistic process that would link estrogen-suppressed bone turnover and decreased DFS among ZA-treated women. Bone metastases in premenopausal women may be less dependent on calcium and on cytokines liberated from bone and the matrix by osteoclasts than those in postmenopausal women. Th is could render the premenopausal women less sensitive to downstream eff ects of ZA that lower local calcium and cytokine levels. Another possibility is that the bone microenvironment in estrogen-exposed women better supports the survival and expansion of disseminated tumor cells in the endosteal niche. Th is idea is supported by the fi ndings that estrogen increases the number and activity of endosteal osteoblasts [22], which are critical mediators of stem cell dormancy and survival [23]; estrogen may thereby impede the ability of ZA to decrease disseminated tumor cells. To better understand whether estrogen and ZA interact at the level of dormancy, it will be important to measure the eff ect of ZA on the ability of estrogen-replete and estrogendeprived endosteum to support cancer. Estrogen defi ciency, on the contrary, might bolster ZA action. Estrogen defi ciency has been linked to increased expression of IL-1 (reviewed in [24]). Th ere is early but intriguing evidence that could link IL-1α to the transition from latency to clinical disease. IL-1 has been shown to mobilize hematopoietic stem cells from the stem/ progenitor niche. IL-1 is a TNFα target, and a recent report demonstrated that TNFα-dependent factors have converted latent breast cancer to symptomatic disease in a mouse model [25]. Both of these factors may activate cancer cells in the bone niche [25]. Conceivably, estrogen defi ciency could thereby unmask latent disseminated breast cancer cells in the bone environment, making them susceptible to death from high local ZA concentrations.

Inhibin activity in bone -role in zoledronic acid effi cacy?
Hormones other than steroids could engender a tumor microenvironment that is less responsive to ZA. Th e gonadal peptides inhibin A and inhibin B, which are highly expressed in premenopausal women but are blocked in postmenopausal or ovarian-suppressed premeno pausal women, appear to sustain bone mass. Loss of inhibin has previously been suggested as a factor in the anti-cancer eff ect of ZA in postmenopausal women [26]. Briefl y, decreased inhibin levels in perimenopause are associated with derepression of follicle-stimulating hormone production, deactivation of activins, and bone loss (reviewed in [27]). In both premenopausal women and postmenopausal women, inhibin levels are inversely correlated with bone turnover, independent of estrogen expression. Th e inhibin-poor environment of postmenopausal women (as in the AZURE, Z-FAST and ZO-FAST trials) or ovarian-suppressed women (as in the ABCSG12 trial) can therefore support the high bone turnover condition in which ZA is most eff ective (as discussed above in the context of estrogen deprivation). One caveat is that inhibin may not be eliminated from the tumor setting by goserelin or menopause because of breast cancer cell expression of inhibin A (in ductal carcinoma in situ) and inhibin B (in primary tumors and metastases) subunits [28]. Gonadal-independent expres sion of inhibins in the local breast cancer micro environment could thus potentially off set the signifi cance of endocrine suppression of inhibins. Clearly, additional studies are needed to understand a potential role for the inhibins in ZA activity.

Zoledronic acid eff ects on angiogenesis
Decreased tumor angiogenesis -associated with decreased endothelial proliferation in response to vascular endothelial growth factor (VEGF) -was observed in ZAtreated mice, and ZA-treated patients have decreased circulating levels of the pro-angiogenic molecule VEGF [29]. Indirect eff ects of ZA on angiogenesis arise from eff ects of ZA on macrophage polarization leading to a decrease in tumor-associated macrophages that promote vascularization and support circulating VEGF levels [30].

Candidate estrogen eff ects opposite to zoledronic acid on angiogenesis
Whereas ZA treatment decreased serum VEGF in vitro and in vivo, estrogen is known to cause transient upregulation of VEGF production in noncancerous cells [31], and to cause sustained low-level VEGF expression by cancer cells themselves in vitro and in vivo [32]. In mouse tumors comprised of species-specifi c breast cancer and stromal cells, Saarinen and colleagues demonstrated estrogen induction of the angiogenic cytokines in the stroma, implicating these as mediators of host angiogenesis [33]. In addition, estrogen has been reported to upregulate α 5 -integrin [34] whereas ZA down regulates these integrins and adhesion in endothelial cells [35].

Eff ects of zoledronic acid on macrophage polarization
While tumor infi ltration by macrophages is common, the function of these macrophages can either be immune suppressive and tumor promoting (M2 phenotype) or be tumor suppressive (M1 phenotype). ZA seems to promote the tumor-suppressive phenotype; for example, it promoted a switch of pro-tumorigenic M2 macrophages in co-culture with prostate cancer cells to M1 polarization [36]. In an erb-B2 mouse model, Coscia and colleagues have shown that the ability of ZA to inhibit cancer was correlated with its ability to impair the recruitment of macrophages into tumors and to support M1 polarization of macrophages in tumors as manifested by decreased IL-10 and increased IFNγ production [30]. Th e contribution of macrophages to the anti-cancer eff ect of bisphos phonates has been reviewed recently [37].

Candidate estrogen and progesterone eff ects opposite to zoledronic acid on macrophage polarization
Estrogen has been posited as supportive of the M2 phenotype [38], consistent with reports that it downregulates the M1-promoting cytokine migration inhibitory factor [39]. Like estrogen, progesterone has been found to promote alternative activation of macrophages [40]. Whereas ZA repolarizes macrophages to nitricoxide-producing M1 macrophages, progesterone has been found to down regulate nitric oxide synthase activity in bone marrow macrophages [41]. Additional studies are necessary to fully understand the potential interaction between sex steroids, eff ect of ZA, and macrophage polarization.

Zoledronic acid eff ects on γδ T cells and natural killer cells
ZA increases the immunogenicity of cancer cells by increasing presentation of the prenyl phosphate antigens isoprenyl pyrophosphate and ApppI (resulting from isoprenyl pyrophosphate-AMP binding) on the cell surface [42]. Prenyl phosphate antigens promoted antitumor immunity by activating the tumor-suppressive γδ T-cell subset. γδ T-cell expansion and activation has been confi rmed in cancer patients after ZA administration [43], leading to a phase I trial of ZA plus IL-2 to augment γδ T-cell activity in women with late-phase breast cancer. In addition to promoting IFNγ production by γδ T cells, ZA has been shown to induce IFNγ production by natural killer (NK) cells [44]. Anti-cancer eff ects of ZA have been eliminated in a mouse breast cancer model when IFNγ was knocked out [30]. Th e relative importance of macrophage-generated, NK cell-generated or γδ T-cell-generated IFNγ in ZA eff ectiveness against cancer is unknown.

Candidate progesterone and estrogen eff ects opposing zoledronic acid on γδ T cells or NK cells
Although limited data are available, there is preclinical evidence that estrogen can increase the growth of ERnegative tumors in immunodefi cient mice by suppressing NK-cell cytotoxicity [45]. Estrogen has also been shown to increase levels of granzyme B inhibitor, leading to resistance of cancer cells to killing by NK cells [46]. Th ere is also increasing evidence for progesterone playing a role in the regulation of γδ T cells. While expression of progesterone receptors on lymphocytes (particularly γδ T cells) has been studied primarily in pregnancy, signal transduction through progesterone receptors has recently been reported in peripheral T cells [47]. It has been hypothesized that responsive T cells could produce progesterone-induced blocking factor, an inhibitor of NK cell function [48]. Although far from established, whether progesterone or other sex hormones could oppose the activation of NK cells by ZA would be interesting.

Inhibition of cell migration by zoledronic acid
ZA decreases the migration of mesenchymal stem cells in vitro and lowers their production of CCL5 chemokine [49]. Mesenchymal stem cells migrate from bone marrow to the primary tumor, where they are induced to produce CCL5 that promotes breast cancer migration and metastases [50] -suggesting that anti-tumor eff ects of ZA could be mediated in part through suppression of mesenchymal stem cell movement and activity. ZA also decreases breast cancer invasion [51] and endothelial cell migration [52].

Promotion of migration by estrogen
In contrast to ZA, estrogen has been noted to increase endothelial cell migration [53]. Estrogen can directly increase cancer cell migration and also can increase the ability of mesenchymal stem cells to promote the migration of ER-positive MCF7 breast cancer cells [54]. Th ese pro-migratory activities would probably be antagonized by ZA, however, because the migration is supported by prenylation of Rho and Rac [55].

Growth factor milieu
Zoledronic acid eff ects on pro-tumorigenic growth factor signaling ZA inhibition of Ras prenylation is compatible with suppression of transduction by proliferative cytokines and mitogens. ZA has been reported to alter breast cancer cell responsiveness to growth factors in vitro. Fromigue and colleagues demonstrated that 1 μM ZA blocked the ability of insulin-like growth factor-1 and insulin-like growth factor-2 to support the survival of breast cancer cells cultured in serum-free medium [56]. Th is blocking could be critical since insulin-like growth factors are among the cytokines that are upregulated in the metastatic milieu. ZA has also been reported to inhibit the production of hepatocyte growth factor by macrophages [57]; given that hepatocyte growth factor supports breast cancer cell invasion, chemoresistance and DNA repair capability, this process could also be involved in modulation of the extra-osseous anti-tumor activity of ZA.

Candidate estrogen eff ects opposite to zoledronic acid on pro-tumorigenic growth factors
Th ere is a well-established crosstalk between estrogen signaling and growth factor pathways, including reports on estrogen-mediated induction of a number of growth factors in peritumoral stroma. For example, estrogen has been noted to induce hepatocyte growth factor secretion by macrophages [58] and mammary fi broblasts [59]. Hepatocyte growth factor signaling through the c-met receptor activates mitogen-activated protein kinases inde pen dently of Ras, and could represent a rescue pathway around Ras inactivation by ZA [60].
Th e Kuperwasser group provides another example of indirect tumor-promoting eff ects of hormones acting on stroma or marrow rather than on cancer cells themselves.
Briefl y, they showed that estrogen can promote growth of ER-negative tumors in a stromal-dependent fashion. Interestingly, enhanced growth of ER-negative breast cancer was transferable with bone marrow from estrogen-treated mice [61]. Th e relevance of these studies to human disease is still somewhat unclear given that the benefi t of hormonal blockade as monotherapy clearly depends on the ER expression in the patient's tumor cells.
It is conceivable that cancer recurs in premenopausal women taking ZA because of enhanced formation of peripheral metastatic niches. Th e establishment of the peripheral metastatic niche has been linked to protumorigenic conversion of stroma resulting from granulin-producing bone marrow cells [62]. Estrogen has been shown to upregulate pro-granulin expression in tumor cells and nontumor cells [63]. Conceivably, estrogen could increase granulin levels in migrating marrow cells to increase visceral metastasis. Further studies to determine the eff ect of ZA and sex hormones on the mobilization and function of granulin-expressing cells could be informative.

Resistance to zoledronic acid
Th ere are limited reports of ZA-resistant cell lines arising from long-term low-dose exposure to ZA. An MCF-7 cell line resistant to ZA exhibited cross-resistance to several chemotherapeutic agents, and expressed an increased Bcl2/Bax ratio and increased ABC transporters BRCP and LRP [64]. Analysis of ZA-resistant osteosarcoma cells arising from culture in low-dose ZA disclosed a farnesyl diphosphate synthase-dependent resistance mechanism in one instance and a heat shock protein-27dependent mecha nism in the other [65,66]. Th ese few reports suggest that intrinsic cancer cell resistance to ZA may be multi factorial. Whether extrinsic resistance to ZA arises through restoration of pro-tumorigenic paracrine or juxtacrine factors in the tumor microenvironment is unknown.

Estrogen modulation of hsp27
Estrogen has been shown to transcriptionally upregulate the hsp27 chaperone in both cancer cells [67] and osteoblasts [68], bolstering cell survival in the presence of apoptotic stimuli. Given the fi nding that hsp27 was required for acquired resistance of osteosarcoma cells to ZA [66], it is possible that heightened hsp27 in premenopausal women contributes to ZA resistance among those not treated with anti-endocrine therapy.

Estrogen and the prenylation pathway
It is worth considering whether estrogenic environments directly interfere with downstream signals of ZA. Dalenc and colleagues showed that neither tamoxifen nor the pure anti-estrogen ICI182780 (fulvestrant) aff ected farnesylation [69]. Th e ability of ZA to interfere with prenylation is therefore unlikely to diff er in estrogen-rich and estrogen-poor environments.

Chemoresistance and integrin-5
No datasets that profi le ZA-sensitive and ZA-resistant stroma or tumors are currently available to analyze for enrichment of estrogen-related pathways linked to ZA resistance. Although knowledge of estrogen-induced changes in stroma that cause chemoresistance is limited, much is known about the stromal role in supporting cancer resistance to chemotherapy. Stromal gene signatures have been identifi ed that predict resistance to neoadjuvant chemotherapy for breast cancer [70]. To interrogate this signature for estrogen-directed transcripts, we conducted gene ontogeny and pathway analysis using GeneGo (Metacore, Inc., Philadelphia, PA, USA) software (RAS, unpublished obser vations). Our analysis indicated the strongest association of this resistance signature with signaling through α 5 -integrins (Z score = 144, P = 10 -84 ). As noted earlier, integrin 5 is downregulated by ZA and transcriptionally upregulated by estrogen.

Conclusion and questions for the future
While the mechanism of action is not yet known, adjuvant treatment with ZA does appear to improve DFS and overall survival in early-stage breast cancer patients with ovarian suppression. ZA eff ectiveness as an antitumor (as well as anti-osteolytic) agent could lead to broader incorporation of ZA (and next-generation bisphosphonates) in clinical practice. Optimal use of ZA in cancer will depend on identifi cation of patient subgroups most likely to benefi t, on delineation of mechanisms of ZA anti-tumor action that are consistent with its pharmacokinetics, and on identifi cation of agents that can be optimally combined with ZA and/or can overcome resistance of cancers to ZA action. Progress will depend on correlated preclinical and clinical studies that uncover whether previously reported eff ects of ZA (for example, on γδ T-cell expansion, on circulating VEGF) diff er as a function of hormonal status.
We note that there is support in the literature for complex and multifactorial interaction between ZA and hormones both within and outside the bone environment. Estrogen, acting through osteoblasts, could support a dormant tumor cell niche in bone, enabling disseminated cells to survive high intra-osseous ZA levels. ZA and estrogen could have antagonistic eff ects on cytokine stimulation of angiogenesis, on tumor-promoting or suppressing activation of macrophages, or on the mobilization and function of NK cells acting on tumors.
Th ese examples of candidate ZA/hormone interactions are summarized in Figure 1.
One clinically relevant question is whether hormonesensitive anti-cancer eff ects are specifi c to ZA or are a class eff ect or a general eff ect with bisphosphonates. Th e recently published NSABP-34 trial assigned women with operable primary breast cancer to adjuvant placebo or to the non-nitrogen-containing bisphosphonate clodronate for 3 years. Similar to the AZURE trial, no benefi t in DFS or overall survival was detected in the clodronate arm, whereas treatment with clodronate increased DFS and the metastasis free-interval in the subset of women aged over 50 [71]. Clodronate lacks the ability of ZA to inhibit the mevalo nate pathway, so this trial could highlight a diff erent, common, endocrine-sensitive eff ect of these bisphospho nates.
Other clinically relevant questions include the following. Do other compounds that suppress bone turnover, such as denosumab, have this same eff ect, despite suppressing bone via a diff erent mechanism? Are there patient-specifi c factors other than hormonal status that predict benefi t from ZA? Do the benefi ts of adding ZA to endocrine therapy in postmenopausal women equal the benefi ts of adding chemotherapy for these women? Is there an additive or synergistic eff ect from ZA, endocrine therapy, and chemotherapy? Answers to these questions should help to integrate therapy with ZA into the standard of care for early-stage breast cancer.