Identification of Luminal Breast Cancers That Establish a Tumor-Supportive Macroenvironment Defined by Proangiogenic Platelets and Bone Marrow–Derived Cells

Breast cancer is a heterogeneous disease that is categorized into molecular and histopathologic subtypes based predominantly on analysis of hormone and growth factor receptors—namely estrogen (ER), progesterone (PR), and HER2/Erbb2 (1). Women with triple-negative breast cancer (TNBC; i.e., ER⁻/PR⁻/HER2⁻) are at the greatest risk of early recurrence (2). Luminal breast cancers (LBC), which often include ER⁺ tumors, are the most prevalent form of breast cancer. These tumors are often differentiated and associated with good prognosis, yet some patients with LBC experience recurrent disease even 15 to 20 years after their initial diagnosis and surgery (3). Although classification into these categories has some correlation with patient outcome, it is difficult to accurately predict which patients will relapse. Furthermore, there is no correlation between molecular classification and patient response to current treatment therapies (4).

In some patients with metastatic breast cancer, tumor cells clearly disseminate before surgery, but remain undetected for protracted periods of time before the patient becomes symptomatic (5). Incipient primary tumors and second primary tumors can also exist in a state of indolence before being detected. For example, autopsy studies of people without a medical history of cancer revealed that indolent cancers are highly prevalent within the general population (6). What causes indolent tumors to erupt into overt disease is unknown, making it difficult to predict which cancer patients are likely to relapse or to benefit from preemptive therapy.

The systemic environment is appreciated as an important determinant of tumor malignancy and progression (7). We previously established that indolent cancer cells (“responders”) that are disseminated to various anatomical locations within host mice can be stimulated to form malignant tumors as a consequence of aggressively growing triple-negative breast tumors (“instigators”) located at distant anatomic sites (8, 9). A growing body of evidence supports the notion that tumors that coexist within a patient who has multiple tumor burden (e.g., multiple disseminated metastases) can interact systemically to modulate overall cancer progression (10). Responding tumor outgrowth occurs as a consequence of systemically acting cytokines and bone marrow–derived cells that are rendered protumorigenic by the instigating triple-negative breast tumors. This cascade of events, termed “systemic instigation,” results in the outgrowth of highly desmoplastic, malignant tumors (8). We designed studies to determine if other breast cancer subtypes use these same mechanisms. A deeper understanding of systemic tumor-promoting processes should improve identification of patients who would benefit from adjuvant therapy.

To understand whether LBCs exert similar protumorigenic systemic effects as instigating TNBC, we injected responding human breast cancer HMLER-HR cells (9, 11) contralaterally to either LBC tumor cells [MCF7Ras (12)], TNBC tumor cells [BPLER (13)], or Matrigel vehicle control in Nude mice, according to our human tumor xenograft protocol (Fig. 1A). We recovered tissue from 3 of the 4 mice that had been injected with responding tumor cells opposite Matrigel control; however, microscopic analysis revealed that only 1 tissue plug contained a small responding tumor (∼10 mg), whereas the other 2 did not form bona fide tumors (Fig. 1B). Responding cells formed tumors in 100% of the mice bearing the systemic environments established by TNBC and LBC (Fig. 1B). Importantly, the resulting responding tumors were formed by the human responder cells that had been injected (Supplementary Fig. S1A).

Responding tumors that were instigated by TNBC displayed no observable necrosis and were moderately mitotic (Fig. 1C and D and Supplementary Fig. S1B and S1C). As we observed previously (8), these responder tumors formed a desmoplastic stroma infiltrated by a-smooth muscle actin (αSMA)–positive myofibroblasts (Supplementary Fig. S1D). In contrast, responding tumors growing in the LBC environment had areas of observable edema and necrosis, and were highly mitotic (Fig. 1C and D and Supplementary Fig. S1B and S1C). These responding tumors were extensively vascularized without forming desmoplastic stroma (Fig. 1C and Supplementary Fig. S1D). These histopathologic phenotypes were consistent with breast adenocarcinomas observed in the clinic (2). The control tumor recovered from the Matrigel environment was comprised of viable responder cells only at the tumor periphery (Fig. 1C and D and Supplementary Fig. S1B–S1D).

The differences in responding tumor histopathology suggested that LBC might use different systemic tumor-promoting mechanisms than TNBC. Indeed, levels of the cytokine osteopontin, an endocrine factor that is necessary for TNBC-dependent systemic instigation (9), were secreted at approximately 450-fold lower levels from the LBC tumor cells than from the TNBC tumor cells (P = 0.002) and were no different than that of the responding tumor cells (Supplementary Fig. S1E).

Tumors that responded to the TNBC environment had significantly higher vessel density than the control tissues (∼2.6-fold); however, those in the LBC environment had higher microvessel density than either control tissues (5.7-fold, P = 0.001) or TNBC-induced tumors (2.2-fold, P = 0.014; Fig. 1E and Supplementary Fig. S5A–S5C). Blood vessels in the LBC-induced responding tumors contained few CD31⁺ cells, were weakly positive for mouse endothelial cell antigen (MECA32), and lacked pericyte coverage, as indicated by the absence of associated αSMA-positive cells (Fig. 1F and Supplementary Fig. S2A) even though peritumoral vasculature stained strongly for MECA32 and αSMA (Supplementary Fig. S2A and S2B). Therefore, we examined responding tumors for the presence of VEGF receptor 2⁺ (VEGFR2⁺) cells, which aid the formation of blood vessels to varying extents under different pathologic conditions (14). In the tumors that responded to LBC-dependent systemic instigation, the vast majority of blood vessels were comprised of VEGFR2⁺ cells (Fig. 1F). Vasculature in the tumors instigated by TNBC was predominantly devoid of VEGFR2⁺ cells (Supplementary Fig. S2C).

VEGFR2⁺ endothelial precursor cells have been shown to originate in the bone marrow (14, 15) and their elevated numbers in the circulation correlate with advanced stage in patients with invasive breast cancer (16). Therefore, we examined recruitment of bone marrow–derived cells (BMDC) into the various responding tumors in Nude mice that had been successfully engrafted with GFP⁺ bone marrow cells (BMC) prepared from eGFPRag1^−/− mice (Supplementary Fig. S3A). We recovered tissue plugs 4 weeks after injection of the responding tumor cells, when the average tissue mass in each group was 10 mg (not shown). In the plugs extracted from sites where responding tumor cells had been injected contralaterally to Matrigel, only approximately 5% of the total cellular portion of these tissues comprised GFP⁺ BMDCs (Fig. 1G). The numbers of BMDCs in these tissues were not significantly different from those of the contralateral Matrigel plug or control lung tissues, which contained approximately 3% GFP⁺ cells (Supplementary Fig. S3B and data not shown). In contrast, BMDCs were incorporated to a significantly greater extent into responder tumors promoted by both TNBC (P = 0.006) and LBC (P = 0.012) instigators; approximately 20% of the total cellular portion of these tumors was comprised of GFP⁺ bone marrow–derived cells (Fig. 1G).

In responding tumors from the LBC environment, VEGFR2⁺ cells comprised approximately 50% of the total number of GFP⁺ BMDCs (Fig. 1H). These numbers represented a 39% increase above those of TNBC-induced tumors and a 67% increase above tissues in the Matrigel environment. In consonance with our earlier report (8), protumorigenic Sca1⁺/cKit⁻ hematopoietic BMDCs were incorporated to a significantly greater extent into responding tumors stimulated by TNBC than those recovered opposite Matrigel (Fig. 1H). The contributions of CD11b⁺/CD45⁺, Sca1⁺/cKit⁺, VEGFR1⁺, and CD31⁺ cells to the total GFP⁺ BMDC population were not significantly different between the cohorts of mice (Fig. 1H).

VEGFR2⁺ cells were also approximately 2.7-fold more abundant in the bone marrow of mice bearing LBC tumors than in those bearing TNBC or Matrigel control (Fig. 1I). The numbers of other BMC populations in the marrow, such as CD11b⁺/CD45⁺ myeloid cells, were not statistically different between groups (Fig. 1I).

Collectively, these data indicated that luminal breast tumors mediated the expansion of VEGFR2⁺ BMCs that were subsequently mobilized to distant responding tumor sites. Within the responding tumor microenvironment VEGFR2⁺ cells contributed, at least in part, to the formation of tumor vasculature.

We analyzed the plasma levels of some common pro- and antiangiogenic circulating cytokines as potential mediators of systemic instigation in the LBC environment but found no statistically significant differences between cohorts (data not shown). Others have shown that proteins, including angiogenic regulators, are enriched several hundred-fold in circulating platelets as compared with the plasma and that platelets are potent mediators of angiogenesis (17–19). We therefore reasoned that platelets were ideal candidates as mediators of systemic instigation.

Responding tumors that formed as a consequence of LBC instigation recruited approximately 3.7-fold more platelets than those stimulated by TNBC (Fig. 2A). Therefore, we analyzed responding tumors for expression of ligands that are known to mediate platelet adhesion, including Collagen IV (ColIV) and CD24. ColIV is a potent chemoattractant that recruits platelets to injured vessels during wound healing or tumor formation (20). CD24 is a cell surface glycoprotein that is used as a surrogate marker for differentiation status of breast and other cancer cells (21) and can bind to the platelet-expressed adhesion molecule, p-selectin (22).

The vast majority of blood vessels within the responding tumors instigated by LBC contained areas of exposed collagen, whereas collagen deposition was predominantly confined to the intratumoral extracellular matrix in responders from the Matrigel and TNBC environments (Fig. 2B). Moreover, carcinoma cells from responding tumors that had been instigated by LBC were highly enriched for cell surface expression of CD24, particularly in areas surrounding blood vessels (Supplementary Fig. S3C), when compared with those injected opposite TNBC or Matrigel control, in which CD24 expression was limited to a few responding cells (Fig. 2C). We observed platelets within these CD24- and ColIV-rich areas of LBC-instigated responding tumors compared with similar areas in control responding tumors in which p-selectin–positive platelets were not obvious (Fig. 2D), even though circulating platelet counts were elevated in both cohorts relative to cancer-free mice (Supplementary Fig. S3D).

These results indicated that different tumor-promoting systemic environments had a profound impact on the ability of tumor cells to recruit platelets.

The fact that platelets had selectively accumulated in responding tumors exposed to the LBC systemic environment suggested that they might play a functional role in tumor promotion. Hence, we analyzed the ability of platelets, prepared from various tumor-bearing mice, to stimulate angiogenesis using a standard in vitro human umbilical vein endothelial cell (HUVEC) assay. Resting platelets harvested from cancer-free mice and from mice bearing responding tumors opposite either Matrigel control or LBC tumors induced capillary tube formation to similar extents (Fig. 3A). ADP-activated platelets from mice bearing responding tumors opposite Matrigel control were unable to stimulate angiogenesis above baseline controls (Fig. 3A). In significant contrast, ADP-activated platelets from mice bearing LBC tumors had approximately 4-fold enhanced angiogenesis-promoting ability (Fig. 3A).

In light of this striking result, we wished to know if the platelet proangiogenic potential was directly imparted by instigating LBC tumors. First, we assessed the proangiogenic capacity of conditioned medium (CM) from cultured cells. As controls, we used medium from resting mouse and human platelets prepared from cancer-free subjects, which minimally promoted capillary tube formation (Fig. 3B). The CM from responding tumor cells did not significantly enhance in vitro angiogenesis to any extent above that of the control CM from resting platelets (Fig. 3B). In sharp contrast, CM from LBC cells significantly enhanced in vitro angiogenesis by approximately 6.5-fold above that of the responder cells or resting platelet controls (Fig. 3B).

Next, we tested whether platelets were capable of taking up proangiogenic factors released by instigating LBC tumor cells. We did so by coculturing naïve, cancer-free mouse or human platelets with either responder or instigator cells in vitro. We interrogated the supernatants from these cocultures for their ability to induce capillary tube formation, reasoning that if platelets absorbed proangiogenic factors from the medium, then supernatants from LBCs would have reduced angiogenic capacity following their exposure to platelets (Fig. 3C). Naïve platelets did not significantly alter the angiogenic ability of the responding tumor cell CM, which was negligible (Fig. 3B and D). Supernatants from LBC instigating cells, which were otherwise highly proangiogenic (Fig. 3B), exhibited approximately 15-fold and approximately 40-fold reductions in angiogenic ability when the LBC cells were cultured with naïve mouse or human platelets, respectively (Fig. 3D). These results were consistent with previous reports (17) and established that both human and mouse platelets were capable of packaging pro- and antiangiogenic factors secreted by LBC tumor cells.

To identify proangiogenic factors carried by platelets during LBC systemic instigation, we conducted a human cytokine array on various platelet lysates. When compared with platelets from mice bearing responding tumors (no systemic instigation), a number of proangiogenic and proinflammatory human cytokines were significantly more concentrated in the platelets from mice bearing LBC instigating tumors, including growth-related oncogene [GRO (P = 0.012)], IFN-g (P = 0.050), interleukin-6 [IL-6 (P = 0.044)], platelet-derived growth factor (PDGF)-BB (P = 0.033), and placental growth factor [PlGF (P = 0.044; Fig. 3E)]. We validated these results by immunostaining platelets prepared from the various cohorts of mice (Supplementary Fig. S4A–S4C). Notably, platelet-derived levels of VEGF and thrombospondin (TSP), the most extensively studied pro-and antiangiogenic cytokines, respectively, were not significantly different between cohorts (Fig. 3E and Supplementary Fig. S4C).

To understand whether some of these cytokines were functioning at responding tumor sites, we examined the activation status of STAT3, which plays an important tumor-supportive role in both breast tumor cells and in the tumor microenvironment (23, 24). STAT3 is a downstream effector of growth factor receptors for cytokines identified in our screen, including IL6 and PDGF (25). Using an antibody specific to the activated, phosphorylated form of both human and mouse STAT3 (p-STAT3), we noted that the levels of p-STAT3 were negligible in control responding tumors opposite Matrigel or TNBC (Fig. 3F). In marked contrast, p-STAT3 staining was abundant in responding tumors that grew contralaterally to LBC tumors, and was localized predominantly to the nucleus of stromal cells within these tumors (Fig. 3F). There was no significant difference in p-STAT3 levels when comparing BMCs from mice bearing the LBC instigating tumors to those of cancer-free controls (Supplementary Fig. S4D), suggesting that the enhancement of STAT3 activity observed in the LBC-bearing mice did not occur in BMCs before their mobilization.

Taken together, these data established that LBC tumors loaded platelets with proinflammatory and proangiogenic factors and provided evidence that these factors were released at distant responding tumors sites. Despite equal concentrations of VEGF and TSP in platelets from both groups, platelets from LBC-bearing hosts had far greater proangiogenic activity, thus underscoring the importance of the complete repertoire of cytokine cargo carried by platelets under different pathologic conditions.

A paradigm of TNBC-mediated systemic instigation is that BMCs (specifically Sca1⁺/cKit⁻ cells) are rendered protumorigenic before mobilization from the marrow into the circulation; hence, when BMCs from hosts bearing TNBCs are admixed with responder cells before injection, the BMCs mimic the effects of the TNBC instigating tumors (8). We therefore tested whether the BMCs that were recruited into responding tumors (Fig. 1G and H) played an active role in the LBC systemic instigation process. To do so, we injected admixtures of responder cells and BMCs prepared from various mice and tested their tumor-promoting ability (Fig. 4A).

BMCs prepared from mice bearing TNBC (BMC-TNBC) were sufficient for responding tumor growth (Fig. 4B). Responding tumors formed in 85% of these mice and tumors were approximately 2.8-fold larger than those that had formed on their own (40% incidence) or that had been admixed with BMCs from cancer-free hosts (68% incidence; Fig. 4B). BMCs from mice bearing LBC tumors (BMC-LBC), however, did not significantly enhance incidence (68%) or mass of responding tumors above that of the cancer-free control BMCs (Fig. 4B). Nevertheless, consistent with the enrichment of CD24 on responder cells in the LBC environment (Fig. 2C and D), CD24 was likewise enriched on the surface of responding tumor cells that had been admixed with the BMC-LBCs (Fig. 4C), and p-selectin–positive platelet aggregates were localized to the CD24-rich areas of these tumors (Fig. 4D). The area covered by CD24⁺ cells in responding tumors admixed with BMCs from control mice was not as extensive as that of tumors admixed with BMC-LBC; consequently, fewer platelet aggregates were observed in the control tumors (Fig. 4C and D).

To understand whether BMCs from LBC tumor-bearing mice directly mediated tumor cell surface enrichment of CD24, we cultured adherent GFP⁺ responding tumor cells with BMCs prepared from various cohorts of mice and analyzed tumor cell expression of CD24 after 4 days by flow cytometry. Responding cells that had been cultured with BMCs harvested from TNBC-bearing mice (BMC-TNBC) displayed an approximately 30% decrease in CD24 expression (Fig. 4E), whereas exposure to BMCs from LBC-bearing mice (BMC-LBC) resulted in an approximately 75% increase in the CD24⁺ responding cell population relative to controls (Fig. 4E).

Hence, BMCs from mice bearing LBC tumors were necessary and sufficient to enrich CD24⁺ responding tumor cells, which recruited platelets. Nevertheless, these events were not sufficient to enhance responding tumor malignancy. We therefore examined the resulting tumors for additional hallmarks of LBC-mediated tumor-promotion. Although VEGFR2+ cells were more abundant in the tumors that had been admixed with BMC-LBC, they did not appear to incorporate into tumor vasculature (Fig. 4F). The numbers of p-STAT3⁺ cells, an indicator for the presence of proangiogenic platelets, were minimal and no different in the BMC-LBC admixed tumors than they were in the BMC-control admixed tumors (Fig. 4G). Consequently, there was no difference in tumor vessel density between the 2 cohorts (Fig. 4H) and in both cases, fewer vessels were apparent than in the tumor-promoting LBC systemic environment (Fig. 1E).

These findings suggested that without the stimulus provided by the LBC tumor, the platelets were not loaded with angiogenesis-promoting cargo and that instigating tumor-educated platelets were crucial for systemic promotion of responding tumor growth.

To understand whether this type of systemic instigation process might reflect real human tumor behavior, we tested 4 different primary human tumor specimens from patients with LBC (hBRCA-LBC 1–4) for their ability to establish a protumorigenic macroenvironment. To do so, we surgically implanted human tumor specimens or Matrigel control plugs subcutaneously into Nude mice contralaterally to responding human cells (Fig. 5A). These patient tumors were diagnosed as invasive ductal carcinoma, grade 2/3, positive for ER and PR, with differing HER2 status, and each human tumor specimen retained its morphological characteristics and cytokeratin expression in the mice (Supplementary Fig. S6A and S6B).

One tumor, hBRCA-LBC1, established a protumorigenic environment that supported growth of highly proliferative responding tumors in 100% (3/3) of the mice (Fig. 5B). This instigating tumor specimen exhibited identical growth kinetics whether it was implanted opposite Matrigel control or opposite the responding tumor cells (Supplementary Fig. S6C). One (1/3) responding tumor was recovered opposite hBRCA-LBC2, none (0/3) were recovered opposite hBRCA-LBC3 or hBRCA-LBC4, and 3 (3/3) tissue plugs were recovered in the Matrigel control conditions.

All of the tumors that responded to hBRCA-LBC1 displayed hallmarks of LBC-mediated tumor promotion, including CD24 enrichment, p-selectin-positive platelet aggregates, p-STAT3 positivity, and vessels marked by exposed collagen and incorporation of VEGFR2⁺ cells when compared with the tumor recovered opposite noninstigating hBRCA-LBC2, in which enrichment of these hallmarks was not observed (Fig. 5C). Consequently, microvessel density in the responding tumors from the instigating hBRCA-LBC1 environment was significantly higher than those of the Matrigel (∼3.4-fold) or noninstigating hBRCA-LBC2 (∼7.2-fold) environments (Fig. 5D and Supplementary Fig. S5D).

These results provided important evidence that primary tumor xenografts could be stratified based on their ability to establish a protumorigenic systemic environment that promoted vascularization and growth of distant disseminated tumors.

Our results suggested that populations of cells that disseminate from a primary tumor in a patient with metastatic disease might respond to systemic signals to convert from a state of indolence to one of overt growth. To test this theory, we selected a tumor specimen from a patient with LBC (hBRCA5, Supplementary Fig. S6D) that remained indolent when implanted into tumor-free control mice (not shown). From this tumor, we prepared organoids (see Methods) and surgically implanted them beneath the skin of Nude mice bearing either instigating LBC tumors or control Matrigel plugs on the contralateral flanks (Fig. 5E).

All of the hBRCA5 organoids implanted into the LBC environment formed aggressively growing tumors that were significantly larger than the control organoids (Fig. 5F) and displayed all of the hallmarks of the LBC-mediated response. Specifically, the instigated hBRCA5 tumors were enriched for CD24, p-STAT3⁺ cells, vessels with exposed collagen, VEGFR2⁺vessel cells, and p-selectin platelet aggregates compared with hBRCA5 organoids in the control environment (Fig. 5C and Supplementary Fig. S6E). Moreover, the instigated tumors had a subtle yet significant increase (∼38%) in vessel density compared with the counterpart control tissues (Fig. 5G).

Clear-cell renal cell carcinoma (cRCC) is typically a highly vascularized cancer. In cRCC patients, levels of circulating VEGFR2⁺ progenitor cells correlate with outcome (26) and tumor cell enrichment of CD24 correlates with reduced progression-free survival (27). We therefore tested the hypothesis that organoids prepared from a nephrectomy surgical specimen, taken from a patient with cRCC, would take advantage of the proangiogenic systemic macroenvironment established by instigating LBC tumors. cRCC organoids were surgically implanted beneath the skin of Nude mice bearing either instigating LBC or control macroenvironments (Fig. 5H). We selected 1 random organoid sample for histology to confirm that all animals received approximately equal portions of tumor tissue (Supplementary Fig. S7A).

cRCC tissues recovered from mice bearing the LBC systemic environment were approximately 2-fold larger in volume and in mass than those that had been implanted opposite Matrigel control (Fig. 5I and Supplementary Fig. S7B). Microscopic examination revealed the presence of tumor cell nodules in 2 of 4 grafts implanted in LBC tumor-bearing mice but in none of the 3 grafts implanted in mice bearing Matrigel plugs (Fig. 5J, top). Moreover, sparse cells with morphological features compatible with cRCC were detected in the other 2 LBC-instigated grafts that did not contain tumor nodules and in one of the grafts implanted opposite Matrigel control (not shown).

Immunohistochemical analysis showed that the tumor cells composing the nodules within the LBC-instigated grafts expressed the cRCC marker Carbonic Anhydrase IX (CAIX; Fig. 5J, inset), and thus retained immunophenotypic features of the fresh tumor sample removed from the patient (Supplementary Fig. S7A). Immunohistochemical labeling of mouse-derived CD34+ cells revealed that one of the tumor nodules displayed a high density of microvessels lined by murine endothelial cells compared with grafts implanted opposite Matrigel control (Fig. 5J, bottom).

Collectively these data indicated that human breast cancer and clear cell renal carcinoma specimens that otherwise did not form successful grafts were able to take advantage of a protumorigenic systemic environment to form vascularized, growing tumors.

To explore therapeutic potential and identify whether platelets were necessary for delivering the protumorigenic instigating stimulus in the LBC macroenvironment, we treated mice with aspirin, which inhibits platelet activity (28). Mice were injected with LBC instigators that were permitted to grow for 4 weeks before initiation of weekly treatments of either aspirin or vehicle control; responders were then injected into these mice 2 days following the first aspirin treatment (Fig. 6A). While responding tumors formed in 90% of the mice treated with vehicle control, only 20% of the mice treated with aspirin developed responding tumors (Fig. 6B). Importantly, aspirin did not significantly affect instigating LBC tumor growth or circulating platelet counts in the 2 cohorts of mice (Supplementary Fig. S8A and S8B).

As expected, responding tumors growing contralaterally to instigating LBC from vehicle-treated mice recruited VEGFR2⁺ cells and were infiltrated with p-STAT3⁺ stromal cells (Fig. 6C). In stark contrast, aspirin treatment completely inhibited the incorporation of VEGFR2⁺ cells into responding tumor sites (Fig. 6C). These tissues were also negative for p-STAT3, suggesting that release of platelet-derived cytokines had not taken place at these tumor sites, as it had in the vehicle-treated controls (Fig. 6C).

Inhibition of responding tumor growth in response to aspirin was not because of a suppression of VEGFR2⁺ cells in the marrow, as the numbers of VEGFR2⁺ cells in the marrows of both vehicle-treated and aspirin-treated mice were approximately 2-fold higher than cancer-free mice and were not significantly different from one another (Fig. 6D). Moreover, aspirin treatment did not affect the ability of BMCs to enrich responding cell CD24 surface expression; tumor cells cultured with BMCs from either vehicle or aspirin-treated mice had an approximately 90% increase in CD24 expression above those exposed to control BMCs from tumor-free mice (Fig. 6E).

These results established that platelets mediated critical steps in the LBC-mediated systemic instigation cascade. Under instigating conditions, aspirin did not affect the instigating tumor or the activity of tumor-supportive VEGFR2⁺ cells in the marrow. Instead, platelet activity manifested most predominantly at the responding tumor site, where platelets were necessary for releasing proangiogenic cytokines and recruiting vessel-forming VEGFR2⁺ cells that facilitated the conversion from indolence to malignancy.

We describe a functional role for the systemic macroenvironment modulated by primary tumors that can ultimately determine the growth and phenotype of secondary tumors (Fig. 7). In the presence of an instigating LBC tumor: (i) circulating platelets are loaded with a repertoire of cytokines that significantly enhances their proangiogenic ability and (ii) the bone marrow is marked by an elevated number of VEGFR2⁺ cells. At the sites where otherwise indolent tumors reside, the proangiogenic platelets accumulate, most likely in response to exposed collagen and tumor cell CD24 presentation. BMCs play a 2-part role: (i) to provide VEGFR2⁺ cells that form the vasculature in response to proangiogenic factors, and (ii) to mediate enrichment of tumor cell CD24, which can serve to recruit the proangiogenic platelets.

The significance of our results using instigating and responding primary human breast tumor specimens is supported by clinical observations that surgical resection of primary tumors improved the survival of women who presented with metastatic breast cancer at the time of diagnosis (29). Analysis of metastatic tumors from breast cancer patients showed that CD24 expression is enhanced on tumor cells at metastatic sites relative to those in the primary tumor (30). CD24 has also been correlated with increased metastatic potential and reduced survival in both breast cancer (31, 32) and cRCC (27) patients. Our study is the first to show that tumor cell enrichment of CD24 has important functional consequences and is directly driven by bone marrow–derived cells, and not just any BMCs, but only those from hosts bearing instigating LBCs.

cRCC is another example of a cancer for which surgical removal of the primary tumor (i.e., cytoreductive nephrectomy) improves patient outcome when conducted before cytokine therapy (33, 34). By showing that human cRCC surgical specimens benefit from the macroenvironment established by instigating LBCs, we do not imply that the mechanisms of systemic instigation apply only to patients with concurrent breast and renal cell carcinoma. Rather, other cancer types, such as cRCC, might operate in a similar fashion to that of instigating LBCs to support the outgrowth of disseminated tumor cells. Indeed, systemic instigation processes might not only apply to the communication system between a primary tumor and its metastases, but between primary tumor foci (i.e., multifocal tumors), multiple primary tumors (i.e., contralateral breast cancer), or different metastatic colonies (7).

It has long been thought that in order for certain tumors to break dormancy, they must undergo an angiogenic switch (reviewed in ref. 35). In these earlier studies, the source of bioavailable proangiogenic factors at the metastatic sites was unclear. We found that platelets provide a rich source of these factors and change the perspective from one in which tumors intrinsically acquire angiogenic potential to one in which angiogenesis is the consequence of processes that are initiated at distant sites. A recent study in a large cohort of ovarian cancer patients established that paraneoplastic thrombocytosis is associated with reduced survival (36). Our study indicates that platelets might not be just an epiphenomenon of malignancy, but actual drivers of malignant progression.

Other studies have shown that interaction with platelets protects metastatic tumor cells from immune surveillance during their voyage in the circulation and helps tumor cells to attach to endothelium upon their arrest at metastatic sites (20). It was recently shown that TGF-β in platelet releasates from cancer-free mice could enhance the metastatic potential of tumor cells (37). Our study expands upon these findings by demonstrating differences in the platelet cytokine repertoire between hosts bearing specific luminal breast tumors and cancer-free hosts, and establishing that platelets continue to promote cancer progression even after tumor cells have disseminated to distant sites. Indeed, platelets provide a major serum source of many proangiogenic proteins in the circulation of patients with breast (38) and colorectal cancer (39). Hence, our finding that both platelet cargo and function are quantitatively different in hosts bearing certain LBCs than they are in cancer-free individuals serves as a prerequisite for analyzing platelets from breast cancer patients.

Therapies targeting VEGF signaling have been less successful in the clinic than first anticipated (40), and approval to use Avastin, a monoclonal antibody against VEGF-A, in women with metastatic breast cancer was recently revoked by the U.S. Food and Drug Administration (41). Our study indicates that the repertoire of cytokines carried by platelets, including, but not limited to, VEGF, mediate proinflammatory and proangiogenic responses in the responding tumor microenvironment to promote tumor progression. Therefore, targeting these other factors, or the platelets that carry them, in the appropriate cohorts of patients should open new therapeutic windows.

Current guidelines suggest prophylactic aspirin use only with respect to cardiovascular protection; however, a role for aspirin in cancer prevention is also suggested from analysis of data from cardiovascular trials (42). Such analyses have established that aspirin use reduces the risk of cancer-related death, particularly from metastatic disease (43–46). Our data support the notion that a subset of breast cancer patients—those with instigating tumors—would benefit from aspirin use and emphasize the need for focused clinical trials because at present, there is not a consensus on whether aspirin use confers a survival benefit to patients with breast cancer. A clear understanding of the molecular and cellular hallmarks that precisely define instigating luminal breast tumors and the protumorigenic environment that they establish has the potential for more accurate selection of patients who would benefit from preemptive treatments, such as aspirin use.

Generation of HMLER-HR, BPLER, and MCF7-Ras human mammary epithelial tumor cells has been previously described (11–13). Expression of cytokeratins and introduced oncogenes was validated for these studies; no additional authentication was conducted by the authors.

Female Nude mice were purchased from Taconic. All experiments were conducted in accordance with regulations of the Children's Hospital Boston Institutional Animal Care and Use Committee (protocol 09-12-1566). Unless otherwise indicated, tumor cells were suspended in 20% Matrigel (BD Biosciences) and injected subcutaneously into nonirradiated mice. Tumors were measured on the flanks of live mice using calipers; volume was calculated as 0.5 × length × (width²).

BMCs were harvested from donor mice by flushing femurs with sterile Hanks' balanced salt solution (HBBS; GIBCO) with penicillin/streptomycin/fungisone. Cells were washed twice with sterile HBBS, dissociated with 18-g needle, and filtered through 70-μm nylon mesh. Bone marrow transplantation was conducted as previously described (ref. 9; see Supplementary Data).

Fresh tissues were digested in 1 mg/mL collagenase A for 1 to 4 hours at 37°C with continuous rotation. Resulting cell suspensions were dispersed with an 18-g needle, washed twice with resuspension buffer (2% heat-inactivated fetal calf serum in sterile HBBS), and filtered through 70-μm nylon mesh. Cells were labeled for flow cytometry by incubation with appropriate antibodies for 30 minutes to 1 hour at 4°C with continuous rotation. Antibodies are listed in Supplementary Data.

Dissected tissues were fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned onto ProbeOn Plus microscope slides (Fisher Scientific) for immunohistochemistry using Vectastain Elite ABC kits (Vector Laboratories, Burlingame CA) as previously described (9) or AlexaFluor fluorescence-conjugated antibodies (Invitrogen). Antibodies are listed in Supplementary Data.

Human and mouse platelets were isolated from whole blood by differential centrifugation as described (ref. 47; detailed in Supplementary Data).

Platelet releasates and cell conditioned media were tested for induction of HUVEC capillary tube formation on Matrigel matrix in vitro using the In Vitro Angiogenesis Assay Kit (Millipore ECM625) according to manufacturer's instructions. Angiogenic ability was quantified by counting branch points, defined as the intersection of 3 or more capillary tubes. Branch points were counted each hour from 4 to 7 hours and data represented as the average number of branch points per sample during the entire experimental time course.

Conditioned media or platelet lysates were tested on Human Angiogenesis Antibody Arrays (RayBiotech) according to the manufacturer's instructions.

Aspirin, 450 mg (Sigma, A2093-100G), was dissolved in 11 mL of dimethyl sulfoxide (DMSO) to make a 41 mg/mL concentrated stock solution. Thirty minutes before injection, 1.6 mL of stock solution was diluted in 11.6 mL of PBS to make a 5 mg/mL injection solution. Vehicle control was made with 1.6 mL of DMSO diluted in 11.6 mL PBS. Mice were injected intraperitoneally with aspirin at 100 mg/kg, once per week. A similar volume:weight ratio of vehicle was administered to control animals.

Primary breast tumors were collected and processed shortly after resection in compliance with a protocol approved by the Brigham and Women's Hospital (IRB 93-085). Each tumor was analyzed for ER/PR/HER2 status. cRCC surgical specimens were obtained with patient consent from the Department of Pathology in compliance with a protocol approved by Brigham and Women's Hospital, Boston MA (DFCI IRB #01-130). All specimens were used without patient identifiers. Tumors were cut to 3- to 4-mm pieces, washed in RPMI, and frozen in RPMI + 10% DMSO. For xenograft studies, tumor specimens were quickly thawed at 37°C, washed 3 times in RPMI, and minced finely into less than 1-mm organoids to ensure tissue homogeneity. Organoids were divided into equal portions, transferred to individual wells of a 96-well plate, covered with 50% Matrigel in RPMI media, and incubated for 10 minutes at 37°C to form cohesive plugs. One organoid plug was selected at random to confirm that samples contained viable tumor cells. Remaining organoid plugs were surgically implanted beneath the skin of Nude mice following sterile surgical procedure as previously described (9).

Data are expressed as mean ± SEM. Data were analyzed by Student t test and were considered statistically significant if P < 0.05.

No potential conflicts of interest were disclosed.

Conception and design: H.S. Kuznetsov, T. Marsh, E.H. Battinelli, S.S. McAllister

Development of methodology: H.S. Kuznetsov, T. Marsh, Z. Castaño, S.S. McAllister

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.A. Markens, A. Greene-Colozzi, A.L. Richardson, S. Signoretti, E.H. Battinelli, S.S. McAllister

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.S. Kuznetsov, T. Marsh, V.E. Brown, S. Signoretti, E.H. Battinelli, S.S. McAllister

Writing, review, and/or revision of the manuscript: H.S. Kuznetsov, T. Marsh, B.A. Markens, S.A. Hay, S. Signoretti, S.S. McAllister

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.S. Kuznetsov, T. Marsh, Z. Castano, S. Hay, B.A. Markens, V.E. Brown, S. Signoretti

Study supervision: H.S. Kuznetsov, T. Marsh, B.A. Markens, E.H. Battinelli, S.S. McAllister

The authors thank Drs. AnhCo Nguyen and Joseph Italiano for helpful reading of the manuscript. The authors thank Rachel Okabe, Mahnaz Paktinat, Ferenc Reinhardt, Amanda Robinson, and Ramya Tadipatri for their technical assistance, and members of the McAllister laboratory for critical discussion.

This work was supported in part by funds from the Dana-Farber/Harvard Breast Cancer SPORE NCI, 2 P50 CA89393-06 (A.L. Richardson); the Dana Farber/Harvard Cancer Center Kidney Cancer SPORE P50 CA101942 (S. Signoretti); NIH NHLBI K08HL097070 and the Malcolm Hecht Jr. Hematology Research Endowment (EMB); the Dana Farber/Harvard Cancer Center Breast Cancer SPORE NCI 5 P30 CA006516-46-2010-01-NN-02, Harvard Stem Cell Institute, and BWH Biomedical Research Institute (S.S. McAllister). S.A. Hay is an American Cancer Society Fuller Research Fellow.