Role of Megakaryocytes in Breast Cancer Metastasis to Bone Free

Bone is a complex and sophisticated organ. Resident marrow cells differentiate either from hematopoietic (HSC) or mesenchymal stem cells (MSC). Chondrocytes, osteoblasts, fibroblasts, adipocytes, endothelial cells, and myocytes originate from MSCs (1). Megakaryocytes differentiate from HSCs in the endosteal niche under the influence of thrombopoietin (TPO) secreted by osteoblasts (2). Mature megakaryocytes are directed to the vascular niche by stromal-derived factor-1 (SDF-1) where they produce and release platelets into the bloodstream. In addition, megakaryocytes play a role in bone metabolism, for example, megakaryocytes inhibit osteoclast function and enhance osteoblast proliferation (3). Bone metastasis is the leading cause of death from breast cancer. Both osteoblasts and osteoclasts are fundamental in the process of metastases growth in the bone. Because thrombocytosis is often seen in many types of advanced cancer, megakaryocytes may assume a much larger role in cancer progression and metastasis than previously known.

“Megakaryocytes join the war on cancer” is the title of a commentary (4) regarding a novel mechanism of host suppression of tumorigenesis via an increase in platelets resulting from an increase in megakaryocytes in the bone marrow of tumor-bearing mice (5). In that study, Lewis lung tumor cells inoculated in the flanks of mice led to an increase in megakaryocytes in the bone marrow and an increase in circulating platelets. The authors concluded that platelet-derived thrombospondin-1 (TSP-1) plays a critical role in suppressing tumor angiogenesis (5). In a different study, Li and colleagues suggest that megakaryocytes are the first line of defense against prostate cancer metastasis to bone, noting a reduction in bone metastasis of PC-3 cells in mice treated with megakaryocyte growth and development factor, TPO, prior to tumor inoculation (6). In contrast, others have suggested that megakaryocytes and platelets act as “primers” of the metastatic niche encouraging cancer colonization (7, 8). Thus, when megakaryocytes join the war, whose side are they on?

While carrying out immunohistochemistry of bone (9), we detected increased numbers of megakaryocytes in the marrow of athymic mice bearing metastatic breast cancer (MDA-MB-231). This observation led us to question whether the megakaryocytes increased in numbers prior to the arrival of the cancer cells to create a hospitable niche or if the megakaryocyte increased in response to cancer cells in the bone? Using a mouse xenograft metastatic model, we found that the increase in megakaryocytes was seen after metastases were evident. However, in a model in which the mammary gland tumor did not metastasize, no increase in the megakaryocytes was observed. In a syngeneic immunocompetent mouse mammary carcinoma metastasis model, 4T1.2 (10), an increase in megakaryocytes was not found in the bone marrow, but rather in the spleen. In Tpo knockout mice, megakaryocytes and platelet counts were reduced by 90% compared with normal mice. When injected with 4T1.2 cancer cells, the Tpo^−/− mice developed metastases more rapidly than either wild-type or heterozygous cohorts.

In order to determine if an increase in megakaryocytes occurred in human metastatic cancer patients, megakaryocytes were counted in bone marrow autopsy samples from patients who died from metastatic breast cancer and from age- and gender-matched subjects whose death was noncancer related. Increased numbers of megakaryocytes were found in six out of eight of the cancer patient samples compared with matched controls.

Taken together, the data suggest that megakaryocytes increase as part of a protective response to metastatic cancer. The loss of megakaryocytes permits a more aggressive metastatic response.

MDA-MB-231-1833-luc (11) is an aggressive bone seeking clone of the MDA-MB-231 human metastatic breast cancer cell line. The cells, a gift, 2010, from Dr. Joan Massagué (Memorial Sloan-Kettering Cancer Center, New York, NY) were grown in DMEM (Corning Cellgro), 10% FBS (PAA Laboratories), and 100 U/mL penicillin/100 μg/mL streptomycin (Corning Cellgro). The mouse mammary carcinoma lines 4T1.2, (bone metastatic) and 67NR (nonmetastatic; ref. 10) were a gift from Dr. Robin Anderson (Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia) in 2008. The nonimmunogenic luciferase tagged 4T1.2 cells (12) were obtained (2009) from Dr. Shoukat Dedhar (British Columbia Cancer Research Centre, Vancouver, British Columbia, Canada), and maintained in DMEM, 10% FBS, 1% nonessential amino acids (Corning Cellgro) penicillin/streptomycin, and 5 μg/mL puromycin; 67NR cells were maintained in DMEM, 10% FBS, 1% NEAA, and penicillin/streptomycin. MEG-01, a human megakaryoblastic line, from Dr. Melissa Kacena (Indiana University, Bloomington, IN) in 2009, were maintained in RPMI1640 supplemented with 10% FBS and penicillin/streptomycin. Murine preosteoblasts, MC3T3-E1, obtained in 2004 from Dr. Norman Karin (Pacific Northwest National Laboratory, Richland, WA) were grown as previously described (13). Primary, authenticated human osteoblasts, NHOst, were purchased from Lonza and grown as recommended. All cell lines were tested and were free of mycoplasma. The MDA-MB-231 and MEG-01 cells both were authenticated by STR profiling (Genetica) most recently in 2016. No universal standards are available for authentication of the murine cell lines (14).

All animal experiments were approved by the Penn State University IACUC, protocol 40832. Animals were anesthetized by inhalation of isoflurane and oxygen. For the xenograft model, female athymic mice, obtained from Charles River Laboratories, Frederick, MD, were inoculated with MDA-MB-231-1833 cells passaged five times without antibiotics. For the experimental metastatic model, 1 × 10⁵ cells were suspended in 100 μL of sterile PBS and injected into the left cardiac ventricle (15, 16) of 8-week-old mice; for the nonmetastatic model, 1 × 10⁶ cells in 25 μL of PBS were inoculated into the fourth mammary gland. Inoculation of PBS alone served as a control. The mice receiving intracardiac injections (six per group) were sacrificed 1, 4, 10, 20, and 30 days postinjection. Mice receiving mammary gland injections (six per group) were sacrificed 4, 14, 24, and 34 days postinjection. Upon sacrifice, blood was collected by cardiac puncture; femurs were removed, fixed in 4% paraformaldehyde, and decalcified in 500 mmol/L EDTA. Bones were paraffin-embedded using standard histologic protocols.

For the syngeneic immunocompetent model, mouse mammary tumor cells, 4T1.2 and 67NR, were injected, 5 × 10⁴ in 25 μL of sterile PBS, into the 4th mammary gland of 13-week-old female BALB/cJ (Jackson Laboratory, Bar Harbor, ME). Sacrifice occurred on days 7, 14, 32, and 40 days postinoculation.

Mice were weighed before cancer cell inoculation and again at sacrifice. Tumor growth was assessed morphometrically using an electronic caliper, and tumor volumes were calculated according to the formula V (mm³) = L (major axis) × W² (minor axis) / 2 (17). On termination, blood, femurs and spleens were collected. Femurs and spleens were fixed and paraffin-embedded.

The MDA-MB-231-1833-luc and 4T1.2-luc cells were engineered to express luciferase for real-time visualization with an IVIS50 Imaging System (Xenogen). The mice were injected with 10 uL/g of body weight of luciferin (15 mg/mL in sterile PBS) 10–15 minutes prior to imaging. Live animal imaging was performed weekly. Total tumor burden was determined by calculating the total amount of photon flux from the luciferase tagged cells within the entire mouse. Living Image(R), version 2.60.1/Igor Pro 4.09A software was used to analyze the IVIS images and to calculate the tumor burden.

Sections of formalin-fixed, decalcified, and paraffin-embedded femurs (10 μm) and fixed and paraffin-embedded spleens (5 μm) were mounted on glass slides, treated with 0.5% trypsin for 15 minutes at 37°C, and stained overnight at 4°C with a primary rabbit polyclonal antimouse von Willebrand (vWF; Abcam) 1:500 in TBS, 1% donkey serum. Endogenous peroxidase was blocked with 3% hydrogen peroxide for 45 minutes at room temperature and the sections were further stained with secondary biotinylated donkey anti-rabbit (Abcam) 1:1,000 in TBS, 1% donkey serum for 2 hours at room temperature. Immunostaining was completed using Vectastain Elite ABC reagent and ImmPACT 3,3′-Diaminobenzidine Kit (Vector Laboratories, Burlingame, CA), and a hematoxylin counterstain. vWF-positive megakaryocytes were counted in 30 fields, ×400 magnification per each femur and spleen with an 20× light microscope.

Human bone marrow samples were obtained from an autopsy set (Penn State Hershey Medical Center) that included eight subjects who died with metastatic breast cancer, and age- and gender-matched individuals who died of noncancer-related causes. Formalin fixed paraffin-embedded sections were stained for vWF using a polyclonal rabbit anti-human vWF (Dako) diluted 1:800 in TBS, 1% donkey serum as the primary antibody. A Keyence BZ-X700 microscope (Keyence Corp.) was employed to image, scan, and quantify the area of each sample. Megakaryocytes were counted and reported as megakaryocytes/mm².

Blood was obtained by cardiac puncture from anesthetized mice. Whole blood (100–500 μL) was collected in EDTA for complete blood counts; the remaining blood was clotted at room temperature for 2 to 3 hours, centrifuged at 400 × g, 10 minutes, and the serum fraction collected and stored at −80°C.

Complete blood cell counts were performed on a Hemavet 950S (Drew Scientific). For some experiments, platelet counts were performed by flow cytometry following the method of Alugupalli and colleagues (18). Briefly, whole blood samples collected in sodium citrate were fluorescently labeled for the platelet markers CD 41 and CD 61. Counts were performed using an Amnis FlowSight Imaging flow cytometer.

SDF-1 and TPO serum levels were determined by ELISA assays (R&D Systems) following the manufacturer's protocol.

C57Bl/6 Tpo knockout mice were created and characterized by de Sauvage and colleagues (19). Frozen embryos were a gift from Genentech. The embryos were re-derived by The Jackson Laboratory and backcrossed with BALB/cJ mice for 10 generations to obtain Tpo heterozygous mice with a BALB/cJ genetic background of >98%. A breeding colony was established at Penn State University to generate Tpo^−/− female mice. Genotype determinations were made from tail snips of 3-week-old pups (Supplementary Fig. S1).

MEG-01 cells were tested for proliferation, adherence, and ploidy in the presence of conditioned media from cancer cells and osteoblasts. Conditioned media were prepared by replacing the cancer cell, MDA-MB-231 or 4T1.2, or osteoblast, murine MC3T3-E1 or human NHOst differentiation medium with serum free RPMI1640, and incubating at 37°C for 24 hours. The media were collected and centrifuged (300 × g, 10 minutes) and stored at −80°C until used. Conditioned media were prepared from nearly confluent cultures of cancer cells, differentiated osteoblasts, or from osteoblasts incubated for 24 hours with 50% conditioned medium from cancer cells. These conditioned media contained secretions from the osteoblasts in response to the cancer cell secretions, and are referred to as “double conditioned medium.” As described in the figure legends, proliferation was based on an increase in viable cell numbers (MTS Promega, CellTiter96 Aqueous One). Adherence of MEG-01 was carried out as described previously (20). The degree of ploidy was measured by flow cytometry following propidium iodide staining. Megakaryocyte progenitors in primary mouse bone marrow were quantified using the MegaCult–C assay system as directed (StemCell Technologies).

The number of animals per group was determined by previous experiments and discussion with a statistician. Data analyses were performed with Graphpad Prism software. One-way and two-way ANOVA with Bonferroni or Tukey multiple comparison tests were used to compare platelet counts, red blood cell (RBC), white blood cell (WBC), WBC populations, megakaryocyte counts, and serum cytokine concentrations. Student t test was utilized to analyze flow cytometry platelet counts and human bone marrow megakaryocyte data.

We noted an increase in megakaryocytes in the femurs of tumor-bearing mice, approximately 1 month after intracardiac inoculation of MDA-MB-231 cells. Using mouse models, we asked when the increase in megakaryocytes occurred during the metastatic process. An increase in megakaryocytes prior to colonization of metastases would support the theory that megakaryocytes prime the premetastatic niche. However, an increase in megakaryocytes after the appearance of metastases would suggest that the cancer cells in the bone marrow affect megakaryopoiesis. In addition, comparison of a metastatic model (intracardiac injection) to a nonmetastatic (mammary gland injection) would indicate whether primary tumors systemically elicited an increase in megakaryocytes.

Mice were injected intracardially with 1833-luc or with PBS (control), and sacrificed, 1, 4, 10, 20, and 30 days postinoculation. Additionally, another group of mice were inoculated in the mammary gland, and sacrificed 4, 14, 24, and 34 days postinoculation. IVIS imaging was performed weekly. Metastases were detected between 4 and 10 days in mice receiving intracardiac injections (Fig. 1A). Mice injected orthotopically showed no signs of metastasis over the course of the study (Fig. 1E). The megakaryocytes in both femurs of each mouse were visualized by immunohistochemical staining of bone sections for von Willebrand factor (Fig. 1B). In the metastatic group, femurs of mice from the 30-day cohort contained about twice the number of megakaryocytes compared with other cohorts (Fig. 1C). The mice sacrificed on day 30 had on average 13.0 (± 2.5 SD) megakaryocytes/field compared with an average of 7.0 (± 1SD) megakaryocytes/field in all other groups. The number of megakaryocytes in the bone marrow increased with time, and correlated with the whole body tumor burden (Fig. 1D). In the nonmetastatic model, there was no difference in megakaryocyte numbers in the animals with or without tumors (Fig. 1F). Although there was a nonsignificant increase in megakaryocyte numbers in the femurs of PBS and tumor injected mice over time, it may be attributed to the aging of the mice.

Platelet counts, performed with an automated counter, indicated variation from mouse to mouse (Fig. 1G). However, on average, the mice injected with 1833 intracardially showed an increase in platelets on day 30, the same time that the megakaryocytes were elevated (Fig. 1G). The mice with mammary gland injections did not have a significantly elevated platelet count (Fig. 1G). These data suggested that tumor cells, once they colonized the marrow, elicited an increase in megakaryocytes. The increase did not appear to be due to a systemic factor released from the primary tumor but rather a response to the presence of the cancer cells in the marrow. Nonetheless, it cannot be ruled out that if the primary tumor had been permitted to grow for a longer time, it might eventually have led to increased megakaryocyte numbers.

In order to determine if the increase in megakaryocytes occurred in mice with an intact immune system, we used a syngeneic model of immunocompetent BALB/cJ mice and murine mammary tumor cells. The megakaryocyte response to nonmetastatic 67NR cells (primary tumors) was compared with that of mice inoculated with 4T1.2 cells (primary and metastatic tumors). Over the course of 6 weeks, six mice per group including a PBS group were sacrificed at 7, 14, 32, and 40 days postinoculation. To monitor metastasis, IVIS imaging was performed weekly (Fig. 2A). Mice were weighed before inoculation and again before sacrifice; tumors were measured immediately prior to sacrifice. In all groups, body weights increased over the course of the study. On day 32, 4T1.2 tumors averaged 154.05 (±43.84 SD) mm³ compared with 67NR tumors that averaged 649.08 (±165.15 SD) mm³. By day 40, 4T1.2 tumors averaged 211.81 (±23.82 SD) mm³ whereas 67NR tumors averaged 833.58 (±177.72 SD) mm³. Although the 67NR tumors grew to nearly four times the volume of the 4T1.2, they did not metastasize, and there was no increase in megakaryocytes.

Because of high levels of G-CSF secreted by 4T1.2 cells, mice with 4T1.2 tumors exhibited splenomegaly (Fig. 2B) and extramedullary hematopoiesis (17). The spleens of mice injected with 4T1.2 cells weighed 0.60 g (±0.5 SD) and measured 2.6 cm (±0.4 SD) in length. By comparison, spleen weights and lengths for PBS and 67NR were (0.09 ± 0.01 g, 1.63 ± 0.05 cm) and (0.09 ± 0.01 g, 1.38 ± 0.08 cm), respectively.

There was no significant increase in megakaryocyte numbers in the marrow of tumor-bearing mice (4T1.2 or 67NR) compared with those inoculated with PBS at any time (Fig. 2C). The average number of megakaryocytes observed across all cohorts was 8.3 (±1.2 SD) megakaryocytes/field. However, there were fewer megakaryocytes in the 4T1.2 injected mice at day 32 (Fig. 2C). This decrease might reflect the large increase in extramedullary megakaryopoiesis in the spleen. In contrast, there was a significant increase in megakaryocytes (Fig. 2D) in the spleens of the 4T1.2 tumor-bearing mice compared with the PBS- and 67NR-inoculated mice. The average numbers of megakaryocytes/field ±SD in the spleens of the 14 day, cancer cell inoculated cohort was 3.5 ± 1.2 compared with 1.2 ± 0.5 in the PBS cohort, and 0.93 ± 0.42 in the 67NR cohort. In the spleens of the 32 days, 4T1.2 cell inoculated cohort the average was 6.7 ± 1.2 compared with 1.6 ± 0.4 in the PBS-injected and 1.15 ± 0.65 in the 67NR inoculated mice. By 40 days after cancer inoculation, the average number of megakaryocytes/field in the spleens of the 4T1.2 mice was 5.1 ± 1.6 compared with 3.0 ± 1 megakaryocytes/field in the PBS controls and 2.31 ± 0.5 in the 67NR-inoculated mice.

There also was a significant increase in platelet counts over time in mice injected with 4T1.2 but not with 67NR or PBS (Fig. 2E) reflecting the increase in megakaryocytes. The increase in platelets in the 4T1.2 inoculated mice was confirmed by a flow cytometric measure of CD41⁺, CD61⁺ markers. At 40 days postinjection, the 4T1.2-bearing mice had on average 1,196 ± 182.3 (SD) platelets × 10³/μL; PBS-injected mice had 384.3 ± 27.0 (SD) platelets × 10³/μL (P < 0.01).

Analysis of the complete blood counts indicated several changes (Fig. 3). The RBCs were significantly decreased in the mice inoculated with 4T1.2 compared with the PBS-injected mice. In contrast, the mice with 67NR had increased RBC counts early on. The decrease with 4T1.2 erythropoiesis was recently reported in another study (21). The 4T1.2 cells also induced a leukemoid reaction (Fig. 3B), resulting in a marked increase in total WBC populations in the 4T1.2 mice compared with control and 67NR inoculated mice (Fig. 3B; ref. 17).There was also a corresponding increase in individual WBC populations (Fig. 3C). The populations of WBCs shifted significantly from leukocytes to granulocytes (mostly neutrophils) in both 4T1.2 and 67NR inoculated mice compared with PBS-treated mice.

Both TPO and SDF-1 are important cytokines for megakaryopoiesis. Decreases in serum TPO and SDF-1 concentrations were noted at 32 and 40 days postinoculation for 4T1.2 mice (Fig. 3D). The 67NR, on the other hand, expressed increased levels. Thus, lower serum TPO levels were associated with metastatic cancer. SDF-1 levels dropped in both the 4T1.2 and 67NR mice (Fig. 3E). The decrease in SDF-1 may be inversely correlated with the increase seen in G-CSF concentrations (22).

In order to further investigate a role for megakaryocytes in cancer metastasis, we utilized Tpo knockout mice, which produce reduced numbers of megakaryocytes (19). Phenotypically, the BALBc/J Tpo knockout strain presented negligible levels of TPO, >90% reduction in megakaryocytes in the bone marrow and spleen, and a nearly 10-fold reduction in platelets compared with wild-type mice. Heterozygous mice showed intermediate levels of these four phenotypic parameters (Table 1).

Tpo^+/+ (wild-type), Tpo^+/− (heterozygotes), and Tpo^−/− (knockout) BALB/cJ mice were inoculated with 4T1.2, and maintained for 4 to 5 weeks to allow for metastasis to occur. We predicted that the Tpo^−/− mice would have fewer metastases than wild type or heterozygotes if the megakaryocytes were positively involved with the metastatic colonization of the bone. Contrary to our prediction, the Tpo^−/− mice developed more aggressive metastases than either heterozygous or wild-type mice (Fig. 4A–C). The IVIS images revealed that the Tpo^−/− mice developed widespread metastases (Fig. 4A and B) by about 33 days. The heterozygotes showed extensive metastases by approximately day 35, whereas the wild-type mice did not show strong evidence of widespread metastases until about day 40. The days of sacrifice were significantly earlier for the Tpo^−/− mice (Fig. 4C) compared with the Tpo^+/− and WT mice. The reduction of the RBCs was greater in the knockout mice at the time of sacrifice compared with the wild-type and heterozygote mice (Fig. 4D). All groups exhibited splenomegaly following injection of 4T1.2 cells.

A count of the megakaryocytes in the femurs and spleens (Fig. 4E and F) revealed a significant increase only in the megakaryocytes in the femurs of Tpo^+/− mice inoculated with the 4T1.2 but not in the other cohorts. In contrast, the increase in megakaryocytes in the spleens was greatest in the wild-type mice. This increase was also seen in the heterozygotes. There were too few megakaryocytes in the knockout mice to be significant. WBC increased in mice injected with wild-type, heterozygotes, and knockout mice (Fig. 4G). However, an analysis of the populations of WBC (Fig. 4H) revealed a shift in lymphocytes to granulocytes in the wild type and in the heterozygotes but not in the knockout mice. The percentage of monocytes remained unchanged among the groups.

The in vivo data suggested that cancer cells that metastasized to the bone produced factors that enhanced megakaryopoiesis. In previous work, we found that osteoblasts in the presence of metastatic cells undergo an inflammatory response and secrete several cytokines, some of which are known to take part in the megakaryopoiesis (13). Thus, we asked if factors produced by the combination of cultures of osteoblasts and cancer cells enhanced steps in megakaryopoieis. We carried out a series of in vitro assays using a human megakaryocyte line, MEG-01, and conditioned media from MDA-MB-231 cancer cells, or from MC3T3-E1 osteoblasts or conditioned medium from MC3T3-E1 incubated with cancer cell conditioned medium (double conditioned medium). We also tested a model of 4T1.2 murine cells and normal human osteoblasts (NHOsts). We asked if the various conditioned media stimulated MEG-01 proliferation/differentiation including adherence and ploidy development. We also assayed for megakaryocyte colony–forming units (CFU-MK).

Proliferation (Fig. 5A) of MEG-01 carried out with 10% FBS or 1% FBS was significantly increased with 10% double conditioned medium, i.e. medium from MC3T3-E1 incubated with MDA-MB-231 conditioned medium. The increase was approximately two- to three-fold above the numbers of the untreated cells. Stimulation of proliferation to a similar extent was also seen with 5% conditioned medium. In contrast the conditioned media from the 4T1.2, NHOst, and combinations did not significantly increase proliferation of the MEG-01 cells (Fig. 5B).

An important stage of megakaryopoiesis is the adherence of megakaryocytes to bone marrow endothelial cells (23). The conditioned media from osteoblasts or the double condition media caused adherence of the MEG-01 cells in the cultures (Fig. 5C). Conditioned medium from the cancer cells alone did not.

Another aspect of the megakaryocyte differentiation is development of polyploidy through endoreplication. Incubating the cells with conditioned media from osteoblasts, cancer cells or double conditioned media did not lead to a significant increase in ploidy (Supplementary Fig. S3).

Finally we asked if the conditioned medium increased the number of CFU-MK in primary bone marrow (Supplementary Fig. S4). None of the conditioned media affected CFU-MK. Only the combination of TPO, IL3, IL6, and IL11 significantly increased CFU-MK.

Is the increase in megakaryocytes seen in mice with metastatic breast cancer also present in the human metastatic patients? It was very difficult to obtain useable clinical bone marrow samples from patients with metastatic bone cancer. However in samples obtained from a matched autopsy set, six out of eight matched pairs indicated an increase in megakaryocytes in the bone marrow of the patient who died of metastatic breast cancer compared with an age and gender matched patient who died of noncancer-related causes (Fig. 6 and Supplementary Table S1). The increase ranged from about 1.4- to 7.5-fold.

Very little research has focused on the interaction of cancer cells and megakaryocytes. Because megakaryocytes are large and reside in the marrow, they are difficult to study. However, megakaryocytes are the sole source of platelets; and platelets have been recognized to be associated with cancer for more than 100 years (see ref. 24). Recent publications concerning the role of megakaryocytes in metastasis offer conflicting theories. In one case, Li and colleagues suggest that megakaryocytes inhibit carcinoma growth in vitro and in vivo (6). When they injected nude mice with TPO to increase megakaryocytes before inoculation of metastatic prostate carcinoma PC-3 cells, a decrease in hind limb tumor burden was seen. They reasoned that because megakaryocytes line the sinusoids of the marrow, they are positioned to be one of the first cells to interact with disseminated tumor cells. An inhibition in metastases suggested that megakaryocytes suppressed PC-3 cells via cell-to-cell contact (6). Zaslavsky and colleagues indicated that platelets and megakaryocyte-derived TSP-1 plays a critical role in suppressing tumor angiogenesis, which limits metastasis in the earliest stages of tumor growth.

In contrast to these protective mechanisms, Psaila and colleagues (7) suggest that megakaryocytes promote metastasis. They imply that the vascular niche is a site for preferential homing and engraftment of malignant cells due in part to the increased localization of megakaryocytes. They point out that megakaryocytes secrete an array of cytokines (7) that stimulate osteoblast proliferation and inhibit osteoclast differentiation. Megakaryocytes also secrete pro- and antiangiogenic cytokines such as VEGF and endostatin (25). In addition, following total body irradiation, megakaryocytes orchestrate regeneration of the marrow compartments (7). The megakaryocytes stimulate osteoblasts to proliferate; thus, they also may stimulate cancer cell growth.

The lack of extensive research in the role of megakaryocytes in metastatic cancer is partly due to the difficulty in obtaining human bone samples that have not been compromised by destructive metastasis. In addition there exists the technical challenge of isolating intact megakaryocytes (7). In this study both a xenograft and a syngeneic model were used to determine the relationship between megakaryocytes and tumor metastasis. We found an increase in megakaryocytes with metastases but not with primary tumors. Because megakaryocytes produce a variety of cytokines and chemokines that affect many cell types and processes, we further asked whether our data supported the idea that megakaryocytes prepare a niche for the cancer cells as suggested by Psaila (7); or acted to protect against cancer cell colonization as suggested by Li (6) and Zaslavsky (5).

In this study, the xenograft intracardiac model provided evidence that metastases colonized the marrow prior to an increase in megakaryocytes. The presence of primary mammary gland tumors for up to 34 days was not sufficient to increase megakaryocytes. We have previously reported that several cytokines including IL6 increase in the bone marrow as part of the osteoblast inflammatory response to cancer cells (16). In addition, IL6 has been found to increase the proliferation of premegakaryocytes (26). Thus, the megakaryocyte pool may expand in response to an increase in IL6 and other cytokines released by the stromal cells in the bone. In vitro experiments carried out with a megakaryocyte line supported the idea that osteoblasts in the presence of cancer cells produce factors that increase megakaryocyte proliferation and adherence, two important steps in megakaryopoiesis. Likely candidates included TPO, IL6, and IL3. The conditioned media did not contain TPO or IL3 as determined by ELISA. IL6 seemed unlikely because murine IL6 does not cross react with human cells. Moreover, addition of neutralizing antibody to IL6 to the cultures did not change proliferation. Also MEG-01 cells secrete IL6, which is likely autoreactive (27).

Using the syngeneic model of spontaneous metastasis from the mammary gland to the bone marrow, we found similar results. In this case, the 67NR tumors were significantly larger than the primary 4T1.2 tumors; nonetheless, there was no increased megakaryopoiesis. Although there was no increase in megakaryocytes in the femurs of mice injected with 4T1.2 compared with those inoculated with 67NR, or PBS, there was a 4-fold increase of megakaryocytes in the spleens, first observed 14 days postinoculation. The spleens also were significantly larger than the control mice by day 14. The in vivo experiments and the in vitro data taken together suggest that megakaryopoiesis increased in response to the presence of metastatic cancer cells in the bone marrow.

We used a Tpo knockout mouse to reduce megakaryocytes (and platelets) speculating that they would have fewer metastases. Contrary to our expectation, TPO knockout BALB/cJ mice injected orthotopically with 4T1.2 cells displayed more aggressive metastasis and a decreased survival compared with wild-type mice injected with 4T1.2. Was this effect due to the reduction of megakaryocytes or to the lack of platelets? There is considerable evidence that platelets aid in metastatic spread through a variety of processes (24). Activation of platelets may allow tumor cells to enter the circulation, and protect tumor cells from immune elimination (24). It is also known that increased platelet levels are a poor prognostic factor for breast cancer metastasis (28), and thromboembolism is a common cause of death in cancer patients (29). There are various reports that platelets protect circulating tumor cells (30). Mezouar and colleagues (31) observed that in an ectopic model of colon cancer, treatment of mice with clopidogrel (Plavix), an antiplatelet drug, decreased the size of the tumors and restored hemostasis. In a syngeneic orthotopic model of pancreatic cancer, clopidogrel significantly inhibited the development of metastases (31). However, the role of megakaryocytes was not investigated.

The results of our study support the idea that circulating platelets are not sufficient to explain an aggressive metastatic phenotype. The finding that Tpo knockout mice with greatly reduced platelet numbers had more aggressive metastasis than the wild type, suggests that other changes in the megakaryocytes or bone marrow were responsible.

One obvious candidate is TPO. The role for TPO in cancer is unknown; however, its role in maintaining HSC quiescence has been established (32). The release of TPO in the endosteal or osteoblastic niche regulates the proliferation of resident HSCs. In Mpl (the receptor for TPO) knockout mice, there is a reduction in the HSC population. The hematopoietic niche (32) is a postulated haven for metastatic cells in the bone (33). Activation of the niche and proliferation of HSC in a Tpo knockout mouse may permit a more fertile environment for metastases. This idea is further supported by the finding that megakaryocytes regulate HSC quiescence. Loss of megakaryocytes activate the HSC niche (34). Another piece of evidence that the HSC niche of the TPO knockout mice was modified, was the lack of a shift from leukocytes to granulocytes following injection of 4T1.2 (Fig. 4G) despite the increase in total WBC (Fig. 4F). We also have a preliminary observation that the bone volume and density in the knockout mice is less than in the wild type.

TPO, together with SCF, GM-CSF, IL1, IL3, IL11, and IL6 are important for megakaryopoiesis (10). Additional chemokines such as LIX, MIP-1, and KC are thought to play a role in early megakaryocyte progenitor proliferation (35). Interestingly, these are produced by osteoblasts in response to cancer (13, 36, 37). An increase in any one of those cytokines in wild-type mice may be due to host cells or be produced by the metastatic cells and could lead to more megakaryocytes. For example, 4T1 tumor cells and MDA-MB-231 constitutively express genes for the myeloid colony-stimulation factors G-CSF, GM-CSF, and M-CSF (17). The secretion of these growth factors contribute to the extramedullary hematopoiesis, splenomegaly, and myelopoiesis noted in the 4T1.2 injected mice. As has been reported in a rat model of metastatic mammary carcinoma, a GM-CSF/IL3-like cytokine, released from the cancer cells led to an increase of megakaryocytes in the femur (7). In addition, a host immune response to the cancer cells took place as several cytokines including IL6 was released into the marrow from resident osteoblasts (16, 38). The increase in IL6 could, in turn, promote megakaryopoiesis. The in vitro studies indicated that IL6 probably did not act alone.

We also observed a significant decrease in SDF-1 (CXCL-12) concentrations in the blood of wild-type BALB/c J mice sacrificed at day 32 and day 40 post 4T1.2 inoculation. SDF-1 and its receptor CXCR4 are known to be important in various human cancer cells and models of metastasis (39, 40). Likewise, CXCR4 is found on Tregs and other bone marrow cells and appears to promote cell mobilization. The disruption of the CXCR4–SDF-1 axis may facilitate mobilization of Treg cells and other CXCR4+ cells into peripheral blood. This disruption may also play a role in cancer cell mobilization, as a decrease in SDF-1 levels late in the experimental time line seemed to coincide with more advanced metastasis.

In humans, G-CSF is often prescribed for patients who undergo chemotherapy; and it causes extramedullary hematopoiesis (41). The resulting megakaryocytes are sometimes mistaken for cancer cells and misdiagnosis occurs (42). From our data it appears that megakaryocytes increase in response to the presence of cancer cells in the bone microenvironment; patients who died from metastatic breast cancer had increased numbers of megakaryocytes in the bone marrow (Fig. 6). Winklemann and colleagues, reported that megakaryocyte ploidy, related to platelet production, was higher in patients who died of metastatic cancer compared with patients who did not (43). Thus, this increase in platelets and megakaryocytes correlated with more metastasis.

Paradoxically, high levels of anti-angiogenic TSP-1 secreted by platelets and megakaryocytes (5) and the cancer cell inhibitory qualities that the megakaryocytes (6) possess, may help to slow down bone metastasis. TPO is used to help replenish platelets in patients undergoing chemotherapy. There is little research that focuses on TPO as a therapeutic. Our finding, that a decrease in megakaryocytes was associated with an increase in metastasis, suggested that megakaryocytes may have a protective role in metastatic cancer to bone. It is possible that in the normal mouse they compete with the tumor cells for the endosteal niche or directly inhibit cancer growth as suggested by Li and colleagues (6).

In conclusion, our data support the hypothesis that in the war on cancer, megakaryocytes battle against metastasis. However, the findings create more questions. Clearly further studies with the Tpo^−/− mice may offer valuable insights into the metastatic process. One preliminary observation that is being pursued is that the bones of the Tpo^−/− mice are less dense than the bones of the wild type. Although metastases were not restricted to bone, but indeed may develop from cancer cells that find haven there.

No potential conflicts of interest were disclosed.

Conception and design: W. Jackson III, D.M. Sosnoski, A.M. Mastro

Development of methodology: W. Jackson III, D.M. Sosnoski, A.M. Mastro

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W. Jackson III, D.M. Sosnoski, S.E. Ohanessian, A. Mobley, K.D. Meisel, A.M. Mastro

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W. Jackson III, D.M. Sosnoski, K.D. Meisel, A.M. Mastro

Writing, review, and/or revision of the manuscript: W. Jackson III, D.M. Sosnoski, S.E. Ohanessian, P. Chandler, A. Mobley, K.D. Meisel, A.M. Mastro

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Jackson III, D.M. Sosnoski, A.M. Mastro

Study supervision: A.M. Mastro

The authors thank Ruth Nissley and John Cantolina of the Huck Institute Microscopy and Cytometry Facility, Penn State University; Yu-Chi Chen, Nick Kendsersky, Johnathan Vicenty, Richa Pursnani, Emily Rutan, William Turbitt, Ashaki Nehisi, Shuleika Lopez Ortiz, Karen Bussard, and Venkatesh Krishnan for technical assistance; Dr. Connie J. Rogers, William Kraemer (Ohio State) and the Penn State Statistical Consulting Center for help with data analysis; and Dr Jake Werner for his help with platelet analysis. We gratefully acknowledge Genentech's generous gift of C57Blk6 Tpo^−/− mouse embryos.

U.S. Army Medical and Materiel Command Breast Cancer Research Program, W81XWH-10-1-0253 to A.M. Mastro, Alfred P. Sloan Foundation Graduate Award to W. Jackson.

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.