Numerous hormonal factors contribute to the lifetime risk of breast cancer development. These include inherited genetic mutations, age of menarche, age of menopause, and parity. Inexplicably, there is evidence indicating that ovariectomy prevents the formation of both estrogen receptor (ER)–positive and ER-negative breast cancers, suggesting that ER-negative breast cancers are dependent on ovarian hormones for their formation. To examine the mechanism(s) by which this may be occurring, we investigated the hypothesis that steroid hormones promote the outgrowth of ER-negative cancers by influencing host cell types distinct from the mammary epithelial cells. We used a novel xenograft mouse model of parturition-induced breast carcinoma formation, in which the tumors that arise following pregnancy lack the expression of nuclear hormone receptors, thereby recapitulating many clinical cases of this disease. Despite lacking ER expression, the tumors arising following pregnancy in this model require circulating estrogens for their formation. Moreover, increasing the levels of circulating estrogens sufficed to promote the formation and progression of ER-negative cancers, which was accompanied by a systemic increase in host angiogenesis and was attendant with the recruitment of bone marrow–derived stromal cells. Furthermore, bone marrow cells from estrogen-treated mice were sufficient to promote tumor growth. These results reveal a novel mechanism by which estrogens promote the growth of ER-negative cancers. [Cancer Res 2007;67(5):2062–71]
The etiology of breast cancer is complex with numerous hormonal factors contributing to the lifetime risk of developing this disease. These include age of menarche, age of menopause, and parity (1, 2). The observation that rates of breast cancer incidence in women, which increase during the premenopausal and perimenopausal phases of life, plateau after menopause (3) provides strong evidence that estrogens are important mediators of breast cancer development. Although it is generally assumed that the growth of estrogen receptor (ER)–negative tumors is not influenced by estrogens, there is evidence to suggest that ovariectomy prevents the formation of both ER-positive and ER-negative breast cancers (4–7), indicating that even ER-negative breast cancers may depend on ovarian hormones for their formation. BRCA1 tumors, the vast majority of which are ER negative, are also effectively prevented by ovary removal (8). These observations suggest that estrogens may play a role in the pathogenesis of breast cancer that operates independently of their well-documented function in binding the ER molecules expressed by breast cancer cells.
Estrogen is primarily produced in the ovaries, with synthesis commencing at puberty and ceasing at menopause. During the follicular phase of the menstrual cycle, circulating estrogen levels are high, contributing to endometrial gland expansion, growth of the underlying uterine stroma, and proliferation and sprouting of associated capillaries (9). In addition, this steroid hormone regulates important biological activities in the cardiovascular, skeletal, immune, and nervous systems (10, 11). For example, in the immune system, estrogen can suppress both T- and B-cell lymphopoiesis in the thymus and bone marrow, whereas in the bone, the ability of estrogen to block the apoptosis of osteoblasts is critical for the prevention of osteoporosis. Hence, estrogen exerts physiologic effects on a variety of tissues throughout the body that are distinct from its effects on reproductive tissues, including the mammary gland.
The mammary gland is a complex tissue, in which the epithelium is embedded within various stromal cells that regulate its proliferation, differentiation, and survival. The progression of normal mammary epithelium toward a neoplastic state depends on the diverse functions of the surrounding stromal cells, including fibroblasts, myofibroblasts, mast cells, macrophages, adipocytes, pericytes, and endothelial cells. These various types of mesenchymal cells seem to be actively recruited into developing tumor masses, where they provide significant contributions to carcinoma cell proliferation, survival, neovascularization, and the acquisition of invasive and metastatic traits (12–14).
One full-term pregnancy reduces the lifetime risk of developing breast cancer (15–18). Notwithstanding this overall protective effect, parous women actually manifest a transient increase in age-adjusted breast cancer incidence in the first 5 years following parturition (15–23). Although recent work has suggested that alterations in extracellular matrix composition may be involved (24), the mechanism(s) underlying these dual, conflicting effects of parity on breast cancer risk remains unclear.
It is widely presumed that the increased risk of developing breast cancer following pregnancy is due to the ability of pregnancy-associated hormones to promote the proliferation of already-initiated target cell populations. In fact, circulating levels of estrogen are at their highest in women during pregnancy (25). However, the majority of breast cancers that develop during this time lack appreciable expression of either the ER or progesterone receptors (PR; refs. 26, 27). This important observation suggests that if hormones like estrogen are involved in promoting breast cancer following pregnancy, they may not be doing so through direct binding to cognate receptors expressed by breast epithelial cells.
We have investigated the contributions of steroid hormones to the growth of cancers that lack ER/PR expression. The present work shows that the pregnancy-associated hormone estrogen can further the growth of tumors, including those lacking the ER, by acting on the stromal cells that are recruited into the tumor mass.
Materials and Methods
Cells, tissue culture, and plasmids. Human breast epithelial cell lines were generated through the introduction of the SV40 large T (LT) antigen, hTERT, and RasV12 as previously described (28). Cells were grown in DMEM:F12 (1:1) supplemented with 5% calf serum, insulin (10 μg/mL), epidermal growth factor (10 ng/mL), and hydrocortisone (1 μg/mL). Two versions of these cells, termed HMLE-Rashi and HMLE-Raslo, differ in their levels of H-Ras protein expression (28). Another variant of these cells, termed HMLE-Her2, was generated in an analogous fashion in which the HER-2/neu oncogene was substituted for the RasV12 protein. PC-3 human prostate cancer cells were purchased from the American Type Culture Collection (Manassas, VA)6
In vivo tumorigenesis and angiogenesis experiments. Animal procedures were done in accordance with protocols approved by the Massachusetts Institute of Technology and Tufts University Institutional Animal Care and Use Committee. All in vivo studies were carried out using immunocompromised nonobese diabetic/severe combined immunodeficient (NOD/SCID) female mice maintained in a specific pathogen-free animal facility (stock no. 001303, Jackson Labs, Bar Harbor, ME). Unlike athymic nude mice, female NOD/SCID mice are fertile and exhibit normal mammary gland development. Involution studies were done by breeding two female mice per one male animal and separating the females after the mice tested positive for vaginal plugs. Following parturition, females were allowed to nurse for 10 days with an equal number of pups per female to permit adequate and uniform lactation for all females. On day 10 or 11 of lactation, all pups were removed from nursing mothers to induce synchronous mammary gland involution. Experiments were initiated 24 h after pup removal.
For tumorigenicity assays, 1 × 106 breast epithelial cells were resuspended in extracellular matrix (Matrigel), diluted 1:3 with culture medium, and injected either s.c. or into the fourth inguinal mammary glands of involuting or age-matched nulliparous female mice.
For hormone studies, slow-release pellets (Innovative Research of America, Sarasota, FL) of 17β-estradiol (1.7, 0.72, or 0.1 mg per pellet), tamoxifen (25 mg per pellet), and RU486l (50 mg per pellet), or carrier compound placebos were introduced s.c. at the time that cells were injected into mice. All hormone pellets were 60-day slow release with biodegradable carrier-binders. Estrogen pellets of mass 0.1, 0.72, and 1.7 mg resulted in circulating plasma levels of 50 to 75, 300 to 400, and >900 pg/mL, respectively. Tamoxifen and RU486 pellets resulted in circulating drug levels of 3 and 2 μg/mL, respectively. Letrozole was kindly provided by Dr. Dean Evans (Novartis Pharma, Basel, Switzerland). Mice received daily s.c. injection of 10 μg letrozole in 100 μL of 0.3% hydroxyl propyl cellulose (in PBS). Control mice received 100-μL injections of vehicle alone.
For angiogenesis studies, 8-week-old female NOD/SCID mice received a 250-μL injection of diluted Matrigel containing 50 ng/mL of basic fibroblasts growth factor (bFGF; Sigma, St. Louis, MO). Mice received either an estrogen pellet (1.7 mg per pellet) or placebo pellet at the time of injection. After 7 days, mice were sacrificed, and the Matrigel plugs, along with blood, bone marrow, mammary glands, and uteri, were isolated.
For bone marrow co-mixing studies, 8-week-old female NOD/SCID mice received a 1.7 mg 17β-estradiol or placebo pellet for 7 days. Treated mice were sacrificed, and femur and tibia bones were flushed with PBS to harvest bone marrow. Cells were washed, and 1 × 106 bone marrow cells were mixed with 2.5 × 105 with HMLE-Raslo (low-ras cells) and resuspended in diluted Matrigel (1:3). The cell mixture was injected s.c. into untreated 8-week-old nulliparous female mice.
Eight-week old NOD/SCID mice either received an estrogen pellet (1.7 mg per pellet) or placebo pellet at the time of injection. After 7 days, mice were sacrificed, and the extracellular matrix plugs, along with blood, bone marrow, mammary glands, and uteri, were isolated.
Bone marrow transplantation. C57BL/6-GFP transgenic mice were crossed onto a Rag-1null background to generate Rag-1null/GFP transgenic mice. Eight- to 10-week-old female mice were sacrificed, and femur, tibia, and humerus bones were flushed with PBS to harvest bone marrow. Cells were washed, and 10 × 106 bone marrow cells were injected into lethally irradiated (3.81-Gy gamma radiation from a cesium source) NOD/SCID mice by retro-orbital plexus administration. Mice were used for experiments 4 to 6 weeks following bone marrow reconstitution.
Peripheral blood was collected from mice 4 weeks following bone marrow transplantation and was treated with RBC lysis buffer (Sigma) following the manufacturer's protocol. The resulting single-cell suspensions were subjected to fluorescence-activated cell sorting analysis. A total of 1 × 105 viable cells, determined by propidium iodide exclusion, were analyzed for each sample.
Histology and immunostaining. For histologic analyses, tumor tissue was removed from animals and immediately fixed in 10% buffered formalin followed by paraffin embedding. Five-micrometer sections were deparaffinized, rehydrated, and subjected to either trichrome staining (to visualize collagen) or H&E staining. Immunohistochemistry was done following deparaffinization and antigen retrieval. Immunocomplexes were visualized by the ABC method (Vector Laboratories, Burlingame, CA). Sections were counterstained with either hematoxylin or methyl green. Immunofluorescence was done on 5-μm frozen sections of tissues that were fixed in 4% paraformaldehyde, saturated with 20% sucrose, and embedded in ornithine carbamyl transferase. Immunocomplexes were visualized with fluorescent avidin-conjugated secondary antibodies (Vector Laboratories). Cell nuclei were visualized by 4′,6-diamidino-2-phenylindole staining.
Tissue sections were incubated with mouse monoclonal, rat or rabbit polyclonal antibodies against SV40 LT antigen (Pab101; Santa Cruz Biotechnology, Santa Cruz, CA), α-smooth muscle actin (NCL-SMA, Novacastra, Burlingame, CA), von Willebrand factor (vWF; DAKO, Carpinteria, CA), ERα (DAKO), ERβ (Santa Cruz Biotechnology), green fluorescent protein (GFP; Abcam, Cambridge, MA), CD34 (Santa Cruz Biotechnology), CD31 (BD Biosciences, San Jose, CA), and human Flt-1 (Santa Cruz Biotechnology).
Angiogenesis quantification. Paraffin sections of extracellular matrix plugs and tumor tissues were subjected to anti-vWF immunohistochemistry to visualize endothelial cells. Using the methodology described in ref. (29), a 49-point Chalkley eyepiece graticule was used to quantify the mean cross-sectional vessel area per section. Mean cross-sectional vessel boundary length was quantified using a Merz straight line–interrupted graticule employed over the same regions (29). For all measurements, at least three to four hotspot fields in at least four independent samples were evaluated. Statistical analyses were done using a two-tailed Student's t test; asterisks denote a t test statistic >0.96, reflecting a statistical significance level of 0.05.
Western blot analysis. Total protein was extracted from cell lines and tumor tissues using an extraction buffer of 50 mmol/L Tris, 100 mmol/L NaCl, 5 mmol/L EDTA, and 1% NP40 detergent in the presence of protease inhibitors (Roche, Indianapolis, IN). Tissues were minced and then homogenized in 5 volumes of this extraction buffer. All tissue debris were removed by centrifugation, and 50 μg of total protein was analyzed by SDS-PAGE analysis. Blots were probe with antibodies against human ERα (Santa Cruz Biotechnology), ERβ (Santa Cruz Biotechnology), and β-actin (Abcam).
Parturition enhances tumor growth, stromalization, and angiogenesis in the mammary gland. Based on the epidemiology described above, we examined whether pregnancy and parturition promotes the growth of mammary carcinomas. To do so, we introduced the potently tumorigenic HMLE-Rashi (28) breast cancer cells into female mice following parturition or into age-matched, nulliparous females. We monitored the animals for subsequent tumor growth, controlling for daily fluctuations in physiology following pregnancy by synchronizing the physiologic states of the mice by inducing simultaneous mammary gland involution (see Materials and Methods). Although HMLE-Rashi mammary tumors formed with high efficiency in both nulliparous mice and in mice injected following pregnancy, the tumors that formed in the latter cohort developed with a significantly shorter latency, arising 2 to 4 weeks before their counterparts implanted in nulliparous hosts. Additionally, the tumors harvested from mice injected following parturition were on average four times larger than their counterparts in nulliparous mice (2,021 versus 525 mg; Fig. 1A).
To test whether the tumor-enhancing effects of pregnancy extended to weakly tumorigenic breast epithelial cells, we repeated the above experiment with HMLE-Raslo cells (28). The HMLE-Raslo cells, as expected, did not form tumors yet remained viable within the mammary gland of nulliparous mice, forming benign epithelial structures (Fig. 1A and B). In contrast, HMLE-Raslo cells injected into mammary glands following pregnancy frequently formed palpable mammary tumors within 8 weeks (8 of 16; Fig. 1A).
The HMLE-Raslo and HMLE-Rashi tumors that developed in mice injected following pregnancy exhibited histologic differences compared with those arising in nulliparous hosts. These differences were most apparent in the structure of the tumor-associated stroma. The abundance of stromal cells was gauged by staining tumor sections for the SV40 LT antigen, which is expressed in the HMLE-Raslo and HMLE-Rashi cells. The HMLE-Rashi mammary tumors that formed in nulliparous mice were poorly differentiated carcinomas with keratinized regions of metaplasia and minimal stromal involvement (9% stroma; Fig. 1C). In contrast, those that arose in mice following pregnancy were significantly more differentiated and included a large proportion of stromal cells within the tumor tissue (44% stroma; Fig. 1C). Tumors arising from HMLE-Raslo cells that were implanted in hosts following pregnancy similarly exhibited a highly stromalized histologic phenotype (Fig. 1C). This enhanced stromal infiltrate was accompanied by a significant increase in neovasculature, as gauged by Chalkley and microvascular density counts (Fig. 1C). The abundant stroma present within tumors growing in post-pregnancy mice exhibited robust desmoplasia, which was characterized by the expression of αSMA and the secretion of large quantities of collagen and other extracellular matrix components (Fig. 1C). In contrast, little collagen deposition or αSMA expression was observed in HMLE-Rashi tumors arising in nulliparous mice (data not shown).
These data indicated that the host environment following pregnancy strongly enhanced tumor formation in the mammary gland; this enhancing effect was observed with both potently tumorigenic and weakly tumorigenic breast epithelial cells. Additionally, tumor formation by both types of tumor cells following pregnancy was associated with a desmoplastic and angiogenic stromal response.
Physiologic alterations associated with pregnancy systemically promote breast tumor formation. The physiologic changes following pregnancy occur both locally within the mammary gland as well as systemically. This raised the question of whether tumor cells growing outside of the mammary gland would also be affected by pregnancy and parturition. To address this possibility, we did experiments analogous to those described above by implanting tumors at s.c. sites located some distance from the mammary gland. For these and subsequent experiments, we used the HMLE-Raslo cells, whose weak tumorigenicity provided a low baseline above which any tumor-enhancing effects could be readily measured.
HMLE-Raslo cells introduced s.c. into mice following pregnancy frequently (19 of 28) formed tumors (Fig. 2A), whereas the same cells introduced into this site in age-matched nulliparous mice failed to do so. The HMLE-Raslo cells injected s.c. following pregnancy formed well-vascularized, highly stromalized tumors, which exhibited a histology similar to that of HMLE-Raslo tumors that developed in the mammary gland following pregnancy (Figs. 1B and 2A). Because implantation within the mammary gland was not required for the enhanced tumorigenesis in these experiments, we concluded that systemic physiologic changes operated to promote tumor growth following pregnancy.
Estrogen systemically promotes the growth of ER-negative cancers. The systemic enhancement of tumor growth described suggested that circulating factors were responsible for the observed tumor formation following pregnancy. Numerous experimental and clinical findings have shown that the female steroid hormones estrogen and progesterone have potent pro-tumorigenic activities (30–32). To determine whether these hormones were involved in promoting tumor growth following pregnancy, we repeated the experiments above, treating cohorts of mice with pharmacologic agents that prevent either estrogen or progesterone function.
Treatment with the PR inhibitor RU486 did not result in tumor inhibition compared with the placebo-treated controls (data not shown); this indicated that progesterone was unlikely to mediate the tumor-promoting effects of pregnancy and parturition. However, administration of letrozole, which prevents aromatase-dependent estrogen synthesis, inhibited s.c. HMLE-Raslo tumor formation in mice following pregnancy when compared with vehicle-treated controls (Fig. 2B). In contrast, treatment with the selective ER modulator tamoxifen failed to inhibit post-pregnancy tumor formation. These observations indicated that the systemic effects of pregnancy and parturition were mediated, at least in part, by estrogen, and that such stimulation could not be blocked by tamoxifen, which of blocks some but not all signaling by of the ER.
To determine if estrogen was sufficient to promote tumor growth in the absence of pregnancy, slow-release pellets of 17β-estradiol or control pellets were implanted into nulliparous mice immediately following s.c. injection HMLE-Raslo cells. Tumor formation occurred in the estrogen-treated animals in a dosage-dependent manner within 6 weeks (Fig. 2C). All mice that received a high dose of estrogen (300–400 pg/mL plasma) developed tumors. At a lower dose of estrogen (50–75 pg/mL plasma), half of the animals developed tumors, exhibiting a frequency and histology similar to that observed for HMLE-Raslo tumors arising in mice following pregnancy (Supplementary Fig. S2A, i–iv). In contrast, placebo-treated mice failed to develop any palpable tumors.
The above experiments indicated that estrogen was necessary for systemic breast tumor formation following pregnancy and indeed could suffice to induce HMLE-Raslo tumor formation in nulliparous mice. Immunoblotting indicated that HMLE-Rashi and HMLE-Raslo cells grown in culture do not express ERα/β (Fig. 2D), and this observation was extended by the finding that these cells do not proliferate in vitro in response to estrogen treatment (data not shown). Likewise, ERα/β was not detectable in lysates from tumors that formed in mice following pregnancy or in estrogen-treated nulliparous mice (Fig. 2D), indicating that tumor formation in vivo was not the result of acquired ER expression. The absence of ERα and ERβ in HMLE-Raslo cells growing in vitro and in vivo made it highly unlikely that the increased tumorigenesis observed was due to the direct effects of estrogen on the HMLE-Raslo cells.
Accordingly, we speculated that estrogen acted on host cells to promote the growth of ER-negative tumors. To determine whether this effect was restricted to the HMLE-Ras mammary epithelial cells, we repeated the above experiments with a human prostate cancer cell line (PC3), which lacks both ERα/β and androgen receptor expression (33). We also examined a breast cancer cell line (HMLE-Neu) that was transformed by ectopically expressed HER-2/neu. When PC3 or HMLE-Neu cells were injected s.c. into estrogen-treated nulliparous mice, there was a marked increase in tumor growth and a reduction in tumor latency relative to placebo-treated controls (Fig. 2E; Supplementary Fig. S1). All estrogen-treated mice developed palpable tumors within 2 weeks following injection, whereas placebo-treated mice started developing palpable tumors at around 4 weeks after injection. These findings extended our earlier observations that the tumor-promoting effects of estrogen did not depend on the expression of ER by tumor cells.
Estrogen and parturition systemically induce angiogenesis independently of tumor growth. Extensive clinical and basic research has implicated estrogen in promoting angiogenesis through several mechanisms, including mitogenic stimulation of differentiated endothelia, mobilization of endothelial precursor cells (EPC) into the peripheral circulation (34), and enhanced resistance of EPC to apoptosis. Estrogen-induced vascularization is known to maintain normal tissue homeostasis in women and to do so independently of its effects on tumor neovascularization (34–36). In particular, estrogen regulates uterine angiogenesis during the follicular phase of the menstrual cycle (35, 37–39). In concordance with this notion, we observed a marked increase (3.5-fold) in vascularization of PC3 tumors derived from 17β-estradiol–treated animals relative to placebo-treated animals (Fig. 3A). Additionally, angiogenesis has been found to be a rate-limiting step for HMLE-Raslo tumor formation (40). Consequently, we investigated whether estrogen could promote tumor angiogenesis by acting on stromal cells of host origin.
To test this notion, we determined whether elevated circulating estrogen levels stimulate angiogenesis systemically. Accordingly, we s.c. neoangiogenesis assays in mice treated with slow-release pellets of either 17β-estradiol or carrier compound placebo. To do so, we s.c. implanted extracellular matrix (ECM) plugs containing bFGF and measured the subsequent in-growth of microvessels into these plugs. Relative to placebo-treated control animals, ECM plugs from 17β-estradiol–treated mice displayed significantly increased neoangiogenesis, as gauged by both Chalkley (2.6-fold) and microvascular density measurements (2.1-fold) done on vWF-stained sections of these ECM plugs (Fig. 3A,, vii–ix). In addition, histologic examination revealed a 2.2-fold increase in stromalization in ECM plug sections from 17β-estradiol–treated mice (Fig. 3A , iv–vi). These data indicated that estrogen enhances systemic angiogenesis independently of any effects that it may have on the tumor-associated stroma and thus on tumor growth.
To determine if the increased vascularization of HMLE-Rashi tumors (Fig. 1C) arising following pregnancy was also the result of a systemic increase in angiogenesis, we did angiogenesis assays in postpartum and age-matched nulliparous mice. Although minimal blood vessel formation was observed in ECM plugs carried by nulliparous mice, ECM plugs present in mice following pregnancy were significantly more vascularized (Chalkley, 2.3-fold; microvascular density, 2.2-fold; Fig. 3B). The increased vascularization of ECM plugs derived from postpartum animals was readily apparent upon gross examination under a dissection microscope (Fig. 3B).
Estrogen stimulates the recruitment of bone marrow–derived cells from the peripheral circulation. During the course of performing these experiments, we noticed that the increase in vascular infiltration in ECM plugs carried by estrogen-treated mice was accompanied by a concomitant increase in ECM plug stromalization with non-vascular cell types (Fig. 3A; Supplementary Fig. S2). Because estrogen has been reported to influence the recruitment of various cell types from the peripheral circulation during vascular repair (34, 41), we examined whether this process was occurring in our experimental model. We therefore repeated the experiments above in chimeric mice whose hematopoietic systems had been reconstituted with GFP-labeled donor marrow (Fig. 4A).
Compared with placebo-treated controls, ECM plugs and uteri harvested from estrogen-treated chimeric mice revealed a greater total GFP fluorescence under an epifluorescence dissection microscope (Supplementary Fig. S3). Over 70% of the total cells within the ECM plug were GFP-positive, bone marrow–derived cells (Fig. 4B). Although we did not observe CD31+ endothelial cells that were also GFP positive within the ECM plugs, we did note a perivascular association of GFP-labeled, bone marrow–derived cells with the CD31+ endothelium (Fig. 4B). The presence of perivascular GFP-labeled is consistent with previous findings that such bone marrow–derived cells are crucial to the formation of normal, well-consolidated microvasculature (42). These results suggest that estrogen stimulates the recruitment of bone marrow–derived circulating cells to sites of neoangiogenesis and is able to do so in the absence of any tumor growth.
Bone marrow–derived stromal cells are recruited to tumors in estrogen-treated mice. The above experiments indicated that enhancement of ECM plug stromalization by estrogen is accompanied by an increased recruitment of bone marrow–derived cells. In light of these observations, we examined whether the enhanced tumor formation in estrogen-treated animals was also accompanied by stromal cell recruitment from the peripheral circulation.
Accordingly, HMLE-Raslo cells that had been labeled with dsRed fluorescent protein were injected s.c. into estrogen-treated mice whose bone marrow had been previously reconstituted with GFP-labeled bone marrow (Fig. 4C). Following the harvest of tumors that developed in these animals, frozen tissue sections were microscopically examined for the presence of bone marrow–derived cells. We observed that ∼90% of the tumor-associated stromal cells were GFP labeled, indicating that a significant portion of the stroma in these tumors derived from cells originating in the bone marrow (Fig. 4C and D). We observed an extensive overlap between αSMA+ myofibroblasts and GFP-positive cells, indicating that the majority of tumor myofibroblasts were of bone marrow origin (Fig. 4C and Supplementary Fig. S4).
To investigate whether endothelial cells were also recruited into these tumors from the circulation, we stained tumor sections for CD31. Although a significant number of CD31-positive blood vessels were visualized in these tumors, colocalization with GFP was rarely observed. However, there was a significant perivascular association of GFP-labeled cells with CD31-positive vasculature (Fig. 4C; Supplementary Fig. S4), as has been previously described (42, 43) and as we observed with the ECM plugs. Staining for the murine CD34 antigen, a marker for circulating EPCs, indicated the occasional presence of EPCs within the tumor stroma, a significant fraction of which were GFP labeled and therefore bone marrow derived (Fig. 4C , v).
These findings indicated that significant numbers of stromal cells in tumors arising in estrogen-treated animals are of bone marrow origin. These included myofibroblasts and perivascular cells and only small numbers of EPC, few of which seemed to differentiate into endothelial cells forming the tumor-associated neovasculature (Supplementary Fig. S4).
Bone marrow cells from estrogen-treated mice promote tumor growth. The experiments above shown that the tumor-promoting effects of estrogen are accompanied by increased stromalization and angiogenesis attendant with the recruitment of bone marrow–derived cells. To determine if bone marrow cells from estrogen-treated mice can directly promote tumor growth, 250,000 HMLE-Raslo cells were mixed with 1 × 106 bone marrow cells isolated from placebo- or 17β-estradiol–treated mice and subsequently injected s.c. into untreated nulliparous mice. Tumors developed in 100% of the mice that received HMLE-Raslo cells co-mixed with bone marrow from estrogen-treated mice, whereas tumors did not develop in any of the mice bearing HMLE-Raslo cells that had been mixed with placebo-treated marrow (Fig. 5A). Similar observations were made from marrow obtained from postpartum mice compared with nulliparous controls (data not shown). Although the HMLE-Raslo cells were injected in nulliparous mice, the s.c. tumors that formed due to admixed bone marrow from 17β-estradiol–treated mice were well vascularized and stromalized, with regions staining positively for αSMA (Fig. 5B). These results suggest that estrogen can influence the activity and constitution of cells within the bone marrow, and that these cells are able to directly participate in and promote tumor growth when present within tumor masses. Because the resulting reconstructed tumors were histologically very similar to those arising in estrogen-treated mice, this suggests that the systemic effects of estrogen derive from the effects of this hormone on the activation and mobilization of cells present in the bone marrow.
The present work describes a novel effect of estrogen on the growth of ER-negative tumors. The fact that estrogen promotes tumorigenesis by several ER-negative breast cancer cell lines and a prostate cancer cell line provides compelling evidence that the presently observed effects occur via the influence of estrogens on the physiology of the tissues of the tumor-bearing host, rather than on the tumor cells themselves. This notion is reinforced by the observation that both parturition and estrogen treatment systemically enhanced angiogenesis in the absence of any neoplastic growth. Studies of estrogen's role in promoting the growth and vascularization of cancers have largely focused on the transcriptional effects of estrogen binding to its receptor in ER-positive mammary and ovarian carcinoma cells. In this report, we describe a second important mechanism by which estrogen promotes the growth of ER-negative and ER-positive cancers (i.e., by systemically enhancing angiogenesis and stromal cell recruitment).
The functions of the ERs in vivo have been revealed through examination of ER knockout (ERKO) mice. Both ERKOα and ERKOβ mice, while viable, display significant reproductive abnormalities as well as defects in the cardiovascular, skeletal, and immune systems (44–47). Significantly, ERKOα mice are also defective in angiogenesis and vascular wound repair (46, 47), implicating estrogen in supporting angiogenesis (34) in the absence of any neoplastic growth. These actions are believed to be mediated, in part, through the action of ERα on the gene encoding vascular endothelial growth factor (VEGF). Indeed, VEGF is a potent angiogenic growth factor that regulates endothelial cell proliferation, migration, and survival, in addition to its ability to mobilize EPC (48). However, we did not observe significant differences in the circulating levels of VEGF following pregnancy (data not shown), suggesting that at least during this time, circulating VEGF is not the prime mediator of the observed increased angiogenesis and tumor growth.
The lack of clinical responses of ER-negative tumors to tamoxifen therapy would seem to be at odds with the notion that estrogen can promote the growth of these cancers. In concordance with known clinical findings, we failed to observe inhibition of tumor growth following pregnancy by tamoxifen. However, we found clear evidence that an aromatase inhibitor was able to block tumor growth in this context. These observations suggest that estrogen indeed stimulates angiogenesis and thus tumor stromal growth, but that tamoxifen does not block those functions of the ER necessary for bone marrow cell recruitment and angiogenesis. Consistent with this notion, previous reports have indicated that tamoxifen does not inhibit the proangiogenic effects of estrogen in the uterus (49, 50).
The present observations indicate that estrogen increases the systemic capacity for angiogenesis, stromalization, and bone marrow cell recruitment, and that this mechanism is in part responsible for promoting tumorigenesis, including the growth of ER-negative tumors. Within the last decade, the development of superior and more efficacious endocrine therapies targeting estrogen biosynthesis, receptor regulation, and function have yielded remarkable successes in the therapy of ER-positive breast cancers, osteoporosis, and cardiovascular disease. There remains a need, however, for effective therapies against ER-negative tumors, which exhibit a poorer prognosis in general than their ER-positive counterparts. An improved understanding of noncanonical consequences of estrogen signaling may facilitate the design and evaluation of novel endocrinal therapies for breast cancer.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
R.A. Weinberg is an American Cancer Society Research Professor and a Daniel K. Ludwig Foundation Research Professor. C. Kuperwasser is an R.B. Sackler Foundation Scholar.
Grant support: Department of Defense grants BC033108 (P.B. Gupta) and BC033108 (C. Kuperwasser), Breast Cancer Research Foundation (R.A. Weinberg), American Cancer Society grant RP85-128-18 (R.A. Weinberg), Jane Coffin Child Fellowship (C. Kuperwasser), and R.B. Sackler Foundation (C. Kuperwasser).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Tony Chavarria for technical assistance with animal experiments; Phil Hinds, Antoine Karnoub, and Scott Dessain for critical reading of the article; and Homayoun Vaziri and Scott Dessain for helpful advice and encouragement during the course of this work.