Pathogenesis of Bone Metastases: Role of Tumor-Related Proteins



INVESTIGATIONS ON the pathogenesis of cancer metastasis to bone have received increased attention by scientists worldwide. This is good news for patients that suffer from the problematic consequences of bone metastases from breast, prostate, thyroid, kidney, or lung cancer, but it also has highlighted the complexity of the disease process and the difficulties in studying the mechanisms of metastasis to bone in vivo and in vitro. Alterations in bone morphology and function caused by metastases are the result of sequential dysregulation of multiple genes, their protein products, and their respective biological functions in both cancer cells and normal cells that interact with the cancer. Knowledge of the pathobiology of the primary tumor, its progression with time to an invasive and metastatic phenotype, vascular spread to bone, entry into the bone microenvironment, growth in bone, and alteration of bone cell function and bone morphology will be required to understand the pathogenesis of bone metastases. This scope of knowledge necessitates the collaborative efforts of scientists with expertise in cancer and bone biology. It is important that we use our current understanding of cancer and bone pathophysiology to develop an accurate picture of the roles different genes and proteins play in the pathogenesis of bone metastases.

There are significant challenges to studies of bone metastasis that may represent part of the reason why this field has received relatively limited attention in the past. These include the physical difficulty of manipulating bone as a tissue, the comparative difficulty of studying bone in vitro, the complexity of obtaining antemortem samples of bone metastases from human patients, and the relative lack of animal models that effectively mimic human disease. In all of these areas, great strides have been made, but much is yet to be accomplished. For example, it has been very difficult to identify animal models that mimic the osteoblastic metastases of prostate cancer in humans.(1–3) In part, this is because of the fact that bone metastases are an uncommon phenomenon in animals with spontaneous cancer in contrast to humans.(4) Mammary neoplasia is prevalent in rodents (mice and rats) and domestic dogs and cats. If these animals develop distant metastases it is usually to the lungs, and bone metastases are rare.(5–7) In the absence of spontaneous animal models, experimental animal models have been developed, such as injection of tumor cells into the left ventricles of rodents or direct injection into bone.(1,8,9)

The definition of bone metastasis can vary depending on the endpoints examined, because the pathogenesis of bone metastasis by cancer is a multistep process. This is an important concept in correlating clinical and experimental data with the overall pathogenesis of bone metastasis. For example, bone metastasis can be defined as radiographic evidence of altered bone morphology using conventional X-ray radiology. This is an easily understood definition of metastasis but represents a late stage of disease progression and likely underestimates the extent of metastatic bone disease. Scintigraphy, computed tomography, magnetic resonance imaging, and histology will show significantly greater numbers of lesions.(1,10) An experimental procedure that results in reduced numbers of metastases solely based on radiological evidence can be interpreted incorrectly as reducing the actual incidence of bone metastasis. In contrast, bone metastasis may be overestimated by sensitive procedures, such as polymerase chain reaction (PCR), which can identify intravascular tumor cells or latent tumor cells in the bone marrow compartment that might not progress to form destructive bone lesions.


Cancers that frequently metastasize to bone sequentially develop altered cellular characteristics during their progression to a metastatic phenotype.(11,12) Progression may involve inhibition of gene expression (such as tumor suppressor genes), activation of gene expression, or alteration in post-translational modifications of proteins. The sequential progression of tumor cells to the metastatic phenotype and some of the genes and proteins involved are summarized in Table 1. Not all cancers develop the phenotype necessary for metastasis to bone. Breast and prostate cancers have a high incidence (60–80%), thyroid, lung, and kidney cancers have a moderate incidence (30–50%), and gastrointestinal cancer has a low incidence (5%) of metastasis to bone.(13) This suggests that different types of cancer have varied patterns of progression resulting in phenotypes that may or may not be permissive for metastasis to bone.

Table Table 1.. Sequential Changesin Cancer Cell Progressionto the Phenotype That Metastasizesto Bone
StageProteins involved or key concepts
Growth of the primary cancerTumor suppressor genes
 Growth factors
 Angiogenic factors
 Regulators of apoptosis
Tissue invasion by tumor cellsCell adhesion molecules
 Matrix metalloproteinases
 Tissue inhibitors of matrix metalloproteinases
Vascular invasion by tumor cellsMatrix metalloproteinases
Transport of tumor cells to the bone marrow microvasculatureAnatomic pattern of blood or lymphatic vessels
Adhesion to bone marrow endothelial cellsIntegrins
Penetration of bone marrow sinusoidsMatrix metalloproteinases
Growth in the bone marrow cavityCytokines
 Growth factors
 Angiogenic factors
 Regulators of apoptosis
 Interactions with bone marrow cells
 Interactions with bone cells
Modification of bone morphologyStimulation of bone resorption
 Bone resorption by tumor cells
 Stimulation or inhibition of bone formation

Growth and differentiation of the primary neoplasm, controlled by growth factors, oncogenes, tumor suppressor genes, and genes that regulate differentiation or apoptosis, will affect the incidence of bone metastasis. The importance of gene expression in the primary tumor was exemplified by a recent investigation by Henderson et al., in which production of parathyroid hormone-related protein (PTHrP) by breast cancer was associated with a less malignant and less invasive phenotype with reduced incidence of bone metastasis.(14) An early step in developing the metastatic phenotype is induction of local tissue and vascular invasion by cancer cells. This represents a transformation of the tumor cell from a relatively immobile cell to a more primitive motile cell that has fewer or modified cell-to-cell junctions, can penetrate the basement membrane, and migrates through connective tissue stroma. This process mimies early development of epithelial tissues in which migration of epithelial cells is necessary for organogenesis. Invasion is associated with altered expression of cell adhesion molecules and tissue inhibitors of matrix metalloproteinases and increased expression of proteolytic enzymes, such as the matrix metalloproteinases.(11)

The anatomical pattern of blood vessels can modify the pattern of cancer metastasis to bone. Most cancers are likely transported to bone by the arterial system after deposition into the venous circulation via the lymphatic and thoracic duct. An exception may be retrograde venous spread of prostate cancer to vertebrae via Batson's paravertebral venous plexus.(15) Bone receives approximately 5–10% of the cardiac output, and bone metastases occur predominantly in areas of active red marrow.(16,17) This may be because of both the anatomical vascular pattern and the “seed and soil” concept of tumor growth (see below). Injection of cancer cells (that have a propensity to metastasize to bone) into the left ventricle of laboratory rodents results in metastases most commonly at the proximal and distal sides of the growth plate.(18) Predisposition of tumor localization at these sites may be partially caused by the vascular anatomy of the growth plate, because growth plates remain open in adult laboratory rodents. There is an intricate arteriolar blood supply to the physis (especially in growing animals, such as laboratory rodents) with capillaries or sinusoids that complete a 180° turn at the growth plate. The turbulence or stasis of blood caused by this pattern may enhance tumor cell localization at this site. In addition, the sinusoids at the growth plate may have increased permeability or porosity at the junction of the primary spongiosa and mineralizing cartilage. This is also the site of predisposition to development of osteomyelitis caused by thromboembolic sepsis in growing animals and humans. In addition to anatomical pattern, adhesion of cancer cells to endothelial cells in bone also may be a requirement for entry into the bone marrow compartment. Adhesion due to integrins or other cell adhesion molecules may selectively enhance cancer cell localization in bone and even at specific sites in bone.(19) It is reasonable to speculate that endothelial cell adhesion molecules could be different at the growth plate compared with other blood vessels in the bone marrow.

After localization and adhesion of cancer cells to blood vessels in the bone marrow, the cells must penetrate the endothelium to enter the bone marrow compartment. This is facilitated by protease secretion by cancer cells, but chemoattraction also may play an important role in the migratory pattern of the cancer cells. Because metastases occur predominantly to red (hematopoietic) marrow, proteins from the marrow or bone cells or proteins released at sites of bone resorption, such as type I collagen, osteocalcin, transforming growth factor β (TGF-β), or insulin-like growth factors, may stimulate migration of the cancer cells into the marrow compartment.(11)

To develop into a definitive bone metastasis, the cancer cells that enter the bone marrow compartment must proliferate, induce angiogenesis, and ultimately alter bone structure. Inhibition of apoptosis also may play a role in tumor growth.(20) Growth in the bone marrow represents a unique physical and paracrine interplay among the cancer cells and the bone microenvironment, including hematopoietic cells, bone marrow stroma, bone cells, and their matrices (seed and soil hypothesis).(21) The seed and soil hypothesis is supported by the fact that certain cancers selectively localize in bone in spite of the fact that bone receives a minor fraction of the cardiac output.

Most cancers (such as breast cancer) grow in the bone marrow compartment in close proximity to bone surfaces and stimulate osteoclastic bone resorption. Osteoclastic bone resorption is induced predominantly by soluble tumor products (such as PTHrP, interleukin-1, and tumor necrosis factor) that interact with bone cells in a paracrine manner.(12) Proteins released from bone during osteolysis (such as TGF-β) may feed back on the tumor cells to further enhance the release of bone resorbing cytokines by cancer cells.(8) In most areas of the bone marrow, cancer cells do not directly contact bone but are separated from bone by adjacent cancer cells, bone-lining cells, osteoblasts, or osteoclasts. In certain tumors, some cancer cells or tumorassociated macrophages adhere to the bone surface and participate in local osteolysis.(22,23) These forms of osteolysis may play a greater role in advanced lesions. In osteolytic lesions, osteoblastic bone formation may be inhibited by paracrine mediators, which can contribute to the reduction of bone at the metastatic site.

Osteoblastic metastases are lesions having both increased bone formation and resorption.(24) When bone formation is greater than resorption, the metastatic lesion is radiodense due to osteosclerosis. Osteoblastic metastases begin in the marrow cavity and are characterized by nonneoplastic woven bone proliferation on endosteal surfaces in association with cancer cells.(1) Most prostate cancer metastases and few breast cancer metastases are osteoblastic. As the lesions progress in size there may be destruction of adjacent cortical bone. The tumor-related factors that induce new bone formation are under investigation and may include endothelin-1, prostate specific antigen, bone morphogenic proteins, urokinase, insulin-like growth factors, osteoprotegerin, or other substances.(24–26) Osteoblastic metastases are rare in spontaneous animal cancers and very difficult to mimic in experimental animal models of bone metastases. The induction of mostly osteolytic bone metastases in laboratory rodents with human carcinoma lines is likely because of the progression of tumor lines away from the osteoblastic phenotype.(1) It is important to evaluate carefully bone morphology at sites of bone metastases in rodent models. Osteolysis associated with tumor growth in the medullary cavity of bones in rodents often results in disruption of the cortex of the bones. Cortical lysis can cause a marked proliferation of periosteal new bone to stabilize the bone (form of callus). This is an expected physiological reaction to cortical lysis and does not represent an osteoblastic reaction, because proliferation of periosteal new bone is not caused by a paracrine action of the cancer cells. However, there have been recent successes in developing rodent models of true osteoblastic metastases (with intramedullary new bone formation) using certain human breast cancer lines injected in the left ventricle of nude mice, implantation of human prostate cancer lines in medullary cavities of nude or severe combined immunodeficient (SCID) mice, and intravenous injection of human prostate cancer cell lines in SCID mice with subcutaneous xenografts of human bone.(2,9,26,27)


Certain cancers that metastasize to bone express proteins originally identified as bone-matrix proteins and include osteopontin, osteonectin, and bone sialoprotein (BSP).(28,29) The definitive role of these proteins in the pathogenesis of bone metastasis is uncertain, but they may enhance tumor growth, protection from immune surveillance, cell-to-cell adhesion, cell-to-bone adhesion, or migration into the bone microenvironment. Koeneman et al. have hypothesized that cancer cells that mimic bone cell function by secreting growth factors, enzymes, or matrix proteins that occur in bone could induce a phenotype that enhances bone metastasis.(28) However, production of bone matrix proteins are not restricted to tumors that metastasize to bone.(30,31) Therefore, expression of these proteins does not necessarily confer a bone metastatic phenotype on cancer cells but may facilitate bone metastasis when combined with other tumor cell features. Conversely, these proteins may function differently in tumor cells that metastasize to bone, such as breast and prostate cells, possibly increasing their potential to metastasize to bone.


Osteopontin is a calcium-binding phosphoprotein secreted by osteoblasts in bone and trophoblasts of the placenta. Osteopontin also is expressed by some osteotropic cancers, including breast and prostate cancer.(32,33) Osteopontin contains a conserved RGD (Arg-Gly-Asp) amino acid sequence and binds to integrins and the ubiquitous CD44 cell adhesion protein (and its variants).(34,35) Production and binding to the surface of cancer cells by osteopontin could enhance bone metastasis by facilitating cell-to-cell or cell-to-matrix adhesion, chemotaxis, and integrin-mediated signal transduction in bone. In addition, osteopontin can promote tumor cell survival by its ability to reduce macrophage nitric oxide-mediated cytotoxicity.(36)


Osteonectin is a multifunctional glycoprotein produced by osteoblasts and other cell types that plays a role in mineralization, cell-to-matrix adhesion, and angiogenesis. Osteonectin is produced by both prostate and breast cancer, in addition to other tumors including malignant melanomas and fibroblasts transformed with v-Ki-ras.(37–39) Osteonectin promotes invasion and metastasis by enhancing matrix metalloproteinase activity in breast and prostate cancers.(29)


BSP is a secreted glycoprotein produced by osteoblasts, is present in bone matrix, and is produced by certain osteotropic cancers, such as breast, prostate, thyroid, and lung cancer.(32,40–43) In contrast, there is little BSP in normal prostate and breast tissue shown by immunohistochemistry.(43,44) BSP has an integrin-binding domain (RGD), which serves as a ligand for osteoclasts and can mediate migration, adhesion, and proliferation by binding to αvβ3 and αvβ5 integrins on breast cancer cells.(45) RGD peptides have been used to block integrin-associated binding of thyroid cancer cells to bone matrix in vitro.(46) It has been reported that human breast cancer cells synthesize BSP, which may promote mineralization of the tumors and render the carcinomas osteotropic.(39,41) Exogenous BSP peptides with the RGD sequence inhibited human breast cancer cell binding to extracellular bone matrix in vitro.(47) BSP, secreted by BSP-producing cancers, can be measured in the serum.(48) Most circulating BSP is protein bound and factor H (a component of complement) has been discovered as one of its binding proteins.(49) Disruption of the BSP-factor H complex is necessary to measure total circulating concentrations of BSP. Serum BSP concentration may serve as a marker for osteotropic cancers.

In this issue of the Journal, Waltregny et al. have investigated the presence of BSP in primary cancers of the prostate and breast and their metastases in visceral organs and bone using immunohistochemistry on banked tissue samples from autopsies of 15 humans. This is a timely investigation because there is a relative lack of studies on the pathogenesis of bone metastasis using human tissue samples because of the infrequency of biopsies or autopsy samples from patients with bone metastases. All of the primary breast tumors that were available for examination (5 of 8 patients) and the bone metastases from the 8 patients were positive for BSP. Most of the primary prostate cancers (5 of 7 patients) and bone metastases (5 of 7 patients) were positive for BSP. These data confirm previous reports of BSP expression by human breast and prostate cancers and support the importance of examining the role of BSP in bone metastasis.(50) Normal bone cells lining endosteal surfaces, including osteoblasts and osteoclasts, also stained positively for BSP.

BSP was present in visceral metastases from multiple organs of the patients. The intensity of the immunohistochemical staining in the primary and metastatic tumors was qualitatively evaluated (0-3+) and nonparametric statistics were used to compare the primary tumors, bone metastases, and visceral metastases. There was no difference in immunohistochemical staining intensity between the primary tumors and their bone metastases; however, the authors did notice that tumor cells from bone metastases that were close to bone were highly positive for BSP. There was a mild yet significant decrease in BSP staining intensity in visceral metastases compared with the bone metastases. These data were used to conclude that BSP production by breast and prostate cancers might play a role in the pathogenesis of bone metastases. Alternate interpretations of the data also are possible, especially in light of the fact that there was no difference in immunohistochemical staining between the primary tumors and their bone metastases. It is feasible that the bone microenvironment enabled the tumor cells to maintain their production of BSP similar to the primary tumor or that the microenvironments of visceral metastases had a negative influence of BSP production by the tumor cells. If the bone microenvironment was necessary for the tumor cells to maintain their BSP expression as suggested, it is unknown what served as the stimulus for BSP production in the primary tumors.

The authors proposed the hypothesis that production of BSP by tumor cells may facilitate bone metastasis by enabling tumor cells to bind to bone matrix directly. In this hypothesis, BSP would serve as a biochemical link between integrin receptors on cancer cells and glutamic acid sites in bone matrix similar to its role in osteoclast binding to bone. However, binding of tumor cells to bone may not be a necessary component for the development of bone metastases for the following reasons. Tumor cells must first localize in the microvasculature of bone and then bind to endothelial cells, penetrate the capillaries or sinusoids, and migrate and proliferate in the bone microenvironment. Most tumor cells in bone metastases are not in contact with the bone surface, but rather are present in the marrow space and are separated from the bone surface by marrow cells or bone lining cells. This is apparent in Fig. 1. The histopathological depiction of bone metastases suggests that entry and proliferation in the marrow space is the most important criteria for development of metastases. In three photomicrographs (2C, 2F, and 4B), the authors indicate that prostate or breast cancer cells are in close approximation to the bone surface and produce BSP. However, Fig. 2C shows that most breast cancer cells in the bone marrow microcompartment are not in contact with bone. In addition, the BSP-positive cells in close association to the bone matrix in Figs. 1C and 1F have a different morphology compared with the tumor cells. They are flatter or more spindle-shaped and the nuclei are smaller. This could be caused by altered tumor cell morphology after binding to the bone surface or it could be because of BSP staining of bone-lining cells separating the cancer cells from the bone surface, as shown in Fig. 1A. Immunohistochemical staining of metastatic tumor cells with epithelial-specific markers (such as keratins) and mesodermal-specific markers (such as vimentin) would be useful to confirm that cells in contact with bone are cancer cells and that there are no osteoblasts or narrow endosteal-lining cells separating cancer cells from the bone surface. Keratin immunohistochemistry was preformed by Lhotàk et al. to show that metastatic human breast cancer cells directly adjacent to bone were associated with matrix metalloproteinase expression and bone resorption.(51) Interestingly, the bone resorption induced by the breast cancer cells resulted in bone surfaces with jagged edges compared with smooth resorption pits induced by osteoclasts.(51)

Bone resorption can promote the development or growth of bone metastases by multiple mechanisms (Fig. 3). Factors released from resorbing bone, such as TGF-β, can serve as chemoattractants or growth factors for cancer cells.(11,52) TGF-β has been reported to be chemotactic, induce matrix metalloproteinase expression, induce adhesion to collagen via integrins, and stimulate PTHrP expression in cancer cells.(11,53) In addition, bone resorption permits expansion of the bone metastasis through the bone marrow, into cortical bone, and eventually outside of the bone. Most of the bone resorption associated with metastatic cancer is performed by osteoclasts and is stimulated by paracrine factors, such as PTHrP and cytokines, produced by the cancer cells.(12) Resorption of bone directly by cancer cells or tumor-associated macrophages also may contribute to bone resorption in advanced lesions, but this mechanism is likely overshadowed by osteoclastic bone resorption.(23,22) Investigations on human bone metastases can be strengthened by pathological and, in some cases, histomorphometric evaluation of the tumor and bone to assess the relative contribution of osteoclast and tumor cell-mediated bone resorption, bone area and the ratio of bone formation and resorption compared with normal bone, and the location (endosteal vs. periosteal) and type of bone formation (woven vs. lamellar) that is occurring at sites of bone metastases.(1–3)

Production of BSP by cancer cells could facilitate binding of the cells to bone matrix (see Fig. 4B) and permit bone dissolution by matrix metalloproteinases secreted by the cancer cells. The immunohistochemical detection of BSP in breast and prostate cells in this study was predominantly in the cytoplasm, but this does not preclude the possibility that BSP also was localized to the surface of the cells. Fedarco and Fisher have reported that serum BSP is bound to the large complement protein factor H.(49) Binding of factor H to BSP on the surface of cancer cells could prevent the proposed mechanism of cancer cell binding to bone matrix. However, coating of tumor cells by BSP and factor H may aid tumor cell survival by blocking immune surveillance and complement-mediated cell lysis.


It is likely that direct binding of metastatic cancer cells to bone matrix is not necessary for development or progression of bone metastases. Most metastatic tumor cells are not in contact with bone but are separated from bone matrix by adjacent tumor cells, hematopoietic cells, bone-lining cells, osteoblasts, or osteoclasts. However, binding of certain tumor cells to bone matrix via cell surface proteins, such as BSP binding to glutamic acid or integrin binding to type I collagen of bone matrix, has the potential to facilitate tumor cell lysis of bone and increase the release of chemoattractants and growth factors for cancer cells from bone. Tumor cell-induced bone resorption may not be required for development of bone metastasis because of the preponderance of osteoclastic bone resorption, at least in the early formation of bone metastases.

Tumor cells can secrete bone matrix proteins, but their role in the pathogenesis of bone metastases remains speculative and invites mechanistic investigations. It is possible that production of bone matrix proteins by cancer cells could enhance development of bone metastases by mechanisms other than facilitating binding of tumor cells to bone matrix. Such mechanisms include inhibition of immune surveillance of tumor cells, growth promotion, selective cell-to-cell adhesion, or migration in the bone marrow microenvironment. The manuscript by Waltregny et al. has added to our understanding of BSP produced by metastatic human cancer, but mechanistic investigations on the role of bone matrix proteins in bone metastasis using primary human tumors will remain challenging because of the inability to modify selectively tumor cell function in vivo in humans. This highlights the importance of developing animal models of bone metastasis that mimic human disease.


This work was supported by grant CA 77911 from the National Institutes of Health, U.S. Public Health Service. I thank Dr. Charles Capen, Dr. Bruce Leroy, Dr. Virgile Richard, Dr. Rani Sellers, Dr. Sarah Tannehill-Gregg, and Dr. Steven Weisbrode from The Ohio State University and Dr. Laurie McCauley from the School of Dentistry, University of Michigan for their review and insightful comments.