METASTASIS TO BONE: CAUSES, CONSEQUENCES AND THERAPEUTIC OPPORTUNITIES

Transcription

1 METASTASIS TO BONE: CAUSES, CONSEQUENCES AND THERAPEUTIC OPPORTUNITIES Gregory R. Mundy The most common human cancers lung, breast and prostate have a great avidity for bone, leading to painful and untreatable consequences. What makes some cancers, but not others, metastasize to bone, and how do they alter its physiology? Some of the molecular mechanisms that are responsible have recently been identified, and provide new molecular targets for drug development. LEUKOERYTHROBLASTIC ANAEMIA A type of anaemia that is associated with cancers that involve the bone marrow, and that is accompanied by increased production of white blood cells. Department of Molecular Medicine, University of Texas Health Science Center, 7703 Floyd Curl Drive, MS#7877, San Antonio, Texas , USA. mundy@uthscsa.edu doi: /nrc867 Most patients with cancer die not because of the tumour in the primary site, but rather because it has spread to other sites. It is difficult to determine, however, precisely how frequently different tumours metastasize to bone. Patients with advanced breast and prostate cancers almost always develop bone metastases, and the chances are high that, in patients who are originally diagnosed with breast or prostate cancers, the bulk of the tumour burden at the time of death will be in bone. How long the patient lives with the tumour is likely to influence whether bone metastases will occur. For example, in patients who quickly succumb to cancer, due to an aggressively growing primary tumour, bone metastases will be relatively uncommon simply because they have not had time to develop. This does not mean that the tumour cells did not have the potential to grow in bone. There are no reliable prevalence figures for people with bone metastases, but estimates can be made. Of the four million people who die in the United States each year, approximately one-quarter die from cancer, and 70% of these have either breast, lung or prostate cancer 1. So, there are probably more than 350,000 people in the United States who die each year with bone metastases, and probably two to three times this number if patients in the European Union and Japan are also included. The number of bone metastases increases when we consider patients who are living with the condition [OK?],as breast and prostate patients cancer frequently live longer than one year. Bone metastases are infrequently silent they are usually associated with severe bone pain, which can be intractable. The mechanisms responsible for bone pain are poorly understood 2 but seem to be a consequence of osteolysis (bone breakdown). There is evidence that bone resorption inhibitors, such as osteoprotegrin (OPG) or bisphosphonates, might be used to alleviate bone pain 3 (BOX 1). Osteolysis is also accompanied by increased bone fragility; susceptibility to fracture is markedly increased, and pathological fractures frequently occur as a consequence of bone metastases. They often occur in load-bearing bones, and are a particular treatment problem when they are present in the neck or shaft of the femur, or in the pelvis. Other consequences of bone metastasis are LEUKOERYTHROBLASTIC ANAEMIA, bone deformity, hypercalcaemia, and nerve-compression syndromes such as spinal-cord compression. (BOX 2). It has been traditional to think of bone metastases as either osteolytic or osteoblastic (FIG. 1), with entirely different factors being responsible for each. From this viewpoint, osteolytic metastases are believed to be caused by osteoclast-activating factors the most important of which might be parathyroid-hormonerelated peptide (PTHrP) and which are released by tumour cells in the bone microenvironment. Osteoblastic metastases, conversely, are believed to be caused by cancer-cell production of factors that stimulate osteoblast proliferation, differentiation and bone formation. We now realize that osteolytic and osteoblas- 584 AUGUST 2002 VOLUME 2

2 Summary Common tumours, such as those of the breast, lung and prostate, frequently metastasize to bone, and in many patients with advanced disease the skeleton is the site of the most significant tumour burden. There are different patterns of bone effects in patients with cancer, ranging from purely or mostly destructive or osteolytic (breast cancer, myeloma) to mostly bone-forming or osteoblastic (prostate cancer). In the case of breast-cancer-causing osteolysis, the main mediator is parathyroidhormone-related peptide (PTHrP), whereas, in osteoblastic lesions, known mediators include endothelin-1 and platelet-derived growth factor. In osteolytic metastasis, there is a vicious cycle in the bone microenvironment, whereby bi-directional interactions between tumour cells and osteoclasts lead to both osteolysis and tumour growth. The molecular mechanisms that are responsible for this vicious cycle are now being clarified and involve tumour-cell production of PTHrP and bone-derived growth factors that are released as a consequence of increased bone resorption. Bisphosphonates interrupt the vicious cycle and cause not only reduction in osteolytic bone lesions, but also decrease the tumour burden in bone. More-effective treatments for interruption of the vicious cycle are now being developed, including specifically neutralizing antibodies to PTHrP and more efficacious osteoclast inhibitors. ALKALINE PHOSPHATASE An enzyme that is found on the cell surface of osteoblasts, is involved in bone mineralization and used as a serum marker of increased osteoblast activity. BONE-SCANNING AGENTS Agents, such as bisphosphonates, which localize to bone, that can be tagged with a isotope that allows their detection by imaging techniques. These are used in clinical medicine to detect sites of active bone turnover or bone disease. OSTEOCALCIN A protein constituent of bone. Its function remains to be clarified, but circulating levels are used as a marker of osteoblast activity. MYELOMA A neoplastic disorder of the plasma cells that is associated with extensive bone destruction and production of large amounts of specific immunoglobulins. tic lesions are two extremes morphological analysis has revealed that, in most patients, bone metastases have both osteolytic and osteoblastic elements. Osteolytic and osteoblastic lesions Although breast and prostate cancer are the two tumour types that most commonly metastasize to bone, the end result of metastasis by each is usually quite different. In breast cancer, bone metastases are predominantly osteolytic. Osteolysis is caused by osteoclast stimulation not by the direct effects of cancer cells on bone 4. Although the dominant lesion is lytic and destructive, there is usually also a local boneformation response, which presumably represents an attempt at bone repair 5. This increase in bone formation in patients with osteolytic lesions is reflected by increased levels of serum ALKALINE PHOSPHATASE a Box 1 Bisphosphonates Bisphosphonates are a class of pyrophosphate analogues that bind with high affinity to mineralized bone surfaces and inhibit osteoclastic bone resorption. They are used extensively to treat patients with diseases of bone loss, such as osteoporosis and Paget s disease, and cancers that cause osteolysis. Common examples of bisphosphonates are pamidronate (Aredia), alendronate (Fosamax), zoledronate (Zometa) and clodronate (Bonefos). The bisphosphonate market for these diseases is over US $1.5 billion dollars annually. The newer bisphosphonates act by inhibiting specific enzymes in the mevalonate pathway of cholesterol biosynthesis in osteoclasts, which, in turn, leads to impaired prenylation of important small GTP-binding proteins such as RHO, and to subsequent changes in the cytoskeletal function that promotes osteoclast apoptosis. marker of osteoblast (bone-forming cell) activity and increased uptake of BONE-SCANNING AGENTS at the site of the lesion. However, despite this secondary increase in local bone formation, the predominant effect is osteolysis. In prostate cancer, alternatively, bone metastases are frequently osteoblastic 6. In these metastases, there is a profound local stimulation of osteoblasts adjacent to the metastatic tumour cells, as measured by alkaline phosphatase and OSTEOCALCIN levels. Up to 25% of patients with bone metastases from breast tumours, however, also have blastic lesions that are similar to those with metastatic prostate cancer, and some patients with prostate cancer have osteolytic lesions that are similar in nature to those seen in patients with metastatic breast cancer. So the concept that there are basically two types of bone metastases is probably too simplistic. The processes of bone resorption and bone formation are almost always linked or coupled although this coupling might be distorted in cancer. Recent observations indicate that there is a spectrum of bone metastasis. At one end, predominantly osteolytic lesions are associated with reduced osteoblast activity that is uncoupled from rates of bone resorption. Bone metastases that are predominantly osteoblastic, alternatively, also have resorptive components. There is much evidence that both resorption and formation are activated in most bone metastases. In addition to the increased uptake of bone-scanning agents at the metastatic site (a reflection of osteoblastic activity), serum markers of bone resorption, such as urinary hydroxyproline, deoxypyridyniline and pyridiniline crosslinks, are also frequently increased Researchers have followed the course of development of osteoblastic metastases from human breast cancer xenografts in nude mice 13. This model allows sequential morphological observations of the metastases information that is impossible to obtain from clinical studies. In this model, metastatic tumours are initially osteolytic characterized by increases in osteoclast activity, which is associated with increased production of bone-resorption markers. This lytic activity is then followed by a wave of bone formation, and osteoclast activity is reduced. The precise molecular mechanisms that are responsible for this switch are not known. These studies, however, have important implications for therapeutic strategies. Treatments that are aimed at inhibiting bone resorption, such as bisphophonates, might be effective not only in treating lesions that are primarily osteolytic, but also in treating osteoblastic metastases if the osteoblastic response is dependent on previous osteoblast activity 12,14. Some types of bone metastases are known to always be osteolytic. MYELOMA cells cause exclusively osteolytic bone lesions, and although myeloma is not strictly a metastasis, there are probably solid tumours that behave in a similar manner. There are also some osteoblastic lesions that seem to have little or no resorptive component. Charhon et al. 6 described some human cancers that cause osteoblastic metastases with NATURE REVIEWS CANCER VOLUME 2 AUGUST helsenet.info

3 Box 2 Hypercalcaemia Hypercalcaemia (increased blood-calcium 1 g/day concentration) is an important complication of Soft-tissue osteolytic bone disease. It occurs relatively frequently in Bone calcium 1000 mg patients with extensive bone destruction, and is Intestine particularly common in breast, lung, renal, ovarian and pancreatic carcinomas, as well as in myeloma. It is 300 mg/day 500 mg/day distressing for the patient, and it must be recognized and Extracellular fluid calcium treated vigorously. 900 mg Hypercalcaemia that occurs in most patients with 150 mg/day 500 mg/day cancer is due to the production of the peptide 850 mg/day Kidney parathyroid hormone-related peptide (PTHrP) by the tumour PTHrP acts on PTH receptors to cause increased bone resorption and increased renal tubular calcium reabsorption 72. Bone destruction is an 150 mg/day important cause of hypercalcaemia, but the important contributing role of renal mechanisms has been under-appreciated. Hypercalcaemia occurs as a consequence of the combination of these effects and by overwhelming of the calcium homeostatic defence mechanisms. This can be appreciated when the calcium homeostasis for a normal adult in zero calcium balance is considered (see figure). The numbers are estimates of the amount of calcium that is exchanged between the extracellular fluid and gut, kidney and bone each day 73. Bisphosphonates a class of drugs that block bone resorption have made an enormous difference to both the frequency and management of hypercalcaemia in patients with cancer. First, they have reduced its occurrence. And second, when hypercalcaemia does occur, it is readily treated, at least initially 74, and nearly all patients show a beneficial response. However, we may have become complacent in the treatment of hypercalcaemia of malignancy 30. Bisphosphonates might have only a transient beneficial effect because renal tubular calcium reabsorption is unaffected by bisphosphonates. This is not apparent in most patients because hypercalcaemia is usually a hallmark of extensive tumour burden and advanced disease, and many patients with hypercalcaemia die within one month of its onset. Neutralizing antibodies to PTHrP have been shown in preclinical studies to be very effective in the treatment of hypercalcaemia 24. no morphological evidence of a resorptive component. So, how do cancer cells metastasize to bone and, when they have reached their destination, how do they set up this cycle of bone formation and destruction? Pathophysiology of bone metastasis The initial steps in the development of bone metastases are similar to those of metastases to any other site. Primary tumour cells invade their surrounding normal tissue by producing proteolytic enzymes, which traverse the walls of small blood vessels in the normal tissue or those induced by the tumour and enter the circulation 15. They then travel to distant organ sites. These events have been described as inefficient, in that many cancer cells do not survive the normal protective host-surveillance mechanisms during these initial stages The cancer cells that do survive can enter the widechannelled sinusoids of the bone-marrow cavity and are positioned to become bone metastases. Cancer cells must possess certain properties for this to occur. They must have the capacity to migrate across the sinusoidal wall, invade the marrow stroma, generate their own blood supply and travel to the endosteal bone surface (FIG. 2). At this site, they stimulate the activity of osteoclasts or osteoblasts, thereby determining whether the subsequent bone metastasis is osteolytic or osteoblastic. Each of these steps involves important molecular interactions between the tumour cells and the normal host cells, and each is a potential target for the development of drugs that are designed to abrogate the metastatic process. a b Figure 1 Types of bone metastasis. Bone metastasis is often classified as either a osteolytic or b osteoblastic, and one of these effects is usually predominant. For example, metastases from breast and lung tumours are generally osteolytic, whereas metastases from prostate cancer are generally osteoblastic. However, most blastic metastases have a resorptive component, and most lytic lesions are accompanied by some attempt, albeit incomplete, of repair or bone formation. 586 AUGUST 2002 VOLUME 2 helsenet.info

4 Primary malignant neoplasm New vessel formation Invasion Embolism Multi-cell aggregates (lymphocytes, platelets) Bone metastases Arrest in distant capillary bed in bone Extravasation Adherence Tumour-cell proliferation Response to microenvironment Endothelial cell Figure 2 The steps involved in tumour-cell metastasis from a primary site to the skeleton. Each of these steps represents a potential therapeutic target to reverse or prevent metastatic bone disease 75. The primary malignant neoplasm promotes new blood-vessel formation, and these blood vessels carry the cancer cells to capillary beds in bone. Aggregates of tumour cells and other blood cells eventually form embolisms that arrest in distant capillaries in bone. These cancer cells can then adhere to the vascular endothelial cells to escape the blood vessels. As they enter the bone, they are exposed to factors of the microenvironment that support growth of metastases. Mechanisms of osteolytic metastasis The mechanisms by which cancer cells cause osteolytic metastasis are gradually being unravelled. In metastatic human breast cancer, the peptide PTHrP is the main mediator of osteoclast activation, and human osteolytic breast cancer cells have been shown to express PTHrP in vivo. PTHrP expression is greater when the tumour cells are present at the metastatic bone site than when they are present in soft-tissue sites or in the breast 19,20. Immunohistochemistry and in situ hybridization experiments have shown that breast cancer metastases in bone express higher levels of PTHrP than cells that have metastasized to soft tissue or that are present in the primary site. This indicates that PTHrP is a specific mediator of osteolysis in metastatic breast cancer, and it is likely to be the mediator of bone destruction in most other osteolytic cancers 21,22. The role of PTHrP in inducing osteolysis is, however, complex. Henderson et al. 23 have recently published a clinical study showing that PTHrP expression by primary tumours is associated with a favourable outcome and less propensity to bone metastasis. Other preclinical and clinical data, however, associate PTHrP production with bone metastatic potential 19,20,24. This could be due to the fact that tumour cells that express high levels of PTHrP are selected for their ability to metastasize to bone, or that the bone microenvironment increases expression of PTHrP from cancer cells that have spread there. Data from Henderson et al. support the latter explanation 23. PTHrP that is expressed by prostate cancer cells has also been reported to have anabolic effects on bone 25. It is possible that this tumour peptide could promote the osteoblastic metastases that are associated with prostate cancers, although there is no direct data to support this. Increased expression of PTHrP is not the only phenotypic change that occurs in breast cancer cells that enter the bone microenvironment. Mutations in genes that encode mutant oestrogen receptors 26, interleukin (IL)-8 (REF. 27) and the receptor for PTH 28 have also been associated with bone metastasis. Tumour cells that reside in different metastatic sites might have subtle differences in phenotype that could affect not just the behaviour of the cells at that site, but also responses to therapy. So, is PTHrP a viable therapeutic target for bone metastases? Osteolysis caused by human breast cancer metastases was shown to be blocked by neutralizing antibodies against PTHrP 24. Furthermore, compounds that specifically decrease PTHrP expression have been shown to inhibit osteolysis caused by human breast cancer cells in vivo 29. Blocking PTHrP might also have other clinical benefits; Ogata et al. 30 have proposed that this peptide might also be associated with cachexia. RANKL and osteolytic bone disease. PTHrP stimulates osteoclast activity by stimulating production of the cytokine RANKL (receptor activator of nuclear factor-κb ligand), which binds and activates its receptor, RANK, which is expressed by osteoclasts (FIG. 3). There is, however, a debate over the exact function of RANKL in the osteolytic bone activity that is associated with human solid cancers or myeloma. RANKL production by stromal cells is a final common mediator of osteoclast activity that is stimulated by several factors. Many researchers have reported that RANKL is expressed by tumour cells in the bone microenvironment, but it is not clear whether the production of RANKL by tumour cells NATURE REVIEWS CANCER VOLUME 2 AUGUST helsenet.info

5 CALVARIA The skull bones. Rodent calvaria are frequently used in organ culture experiments to determine the effects of factors or compounds that stimulate or inhibit bone resorption or bone formation. Tumour cell OPG PTH/PTHrP IL-1 IL-6 IL-11 JNK RANKL RANK IKK + + AP-1 NF-κB Bone resorption Differentiation, fusion, increased survival Osteoclast Osteoblast/ stromal cell Osteoblast progenitor Figure 3 The RANK RANKL system in osteolytic bone metastases. Tumour production of factors such as parathyroid hormone (PTH) or PTH-related pepttide (PTHrP), interleukin (IL)-1, IL-6 and IL-11 stimulate production of receptor activator of nuclear factor-κb (NF-κB) ligand (RANKL) by osteoblasts and stromal cells. Some of these factors (for example, PTHrP) also decrease production of osteoprotegerin (OPG) a decoy receptor that prevents RANKL from binding to its receptor (RANK) on osteoclast progenitor cells. Signalling through RANK in osteoclast progenitors activates transcription factors such as AP1 (activated by JUN N- terminal kinase (JNK), or JUN) and NF-κB (activated by inhibitor of κb kinase (IKK)), leading to the differentiation of osteoclast progenitors into mature osteoclasts. These osteoclasts mediate bone resorption. themselves is sufficient to activate osteolysis 31.Some studies have shown that bone destruction is prevented by treatment with OPG a soluble decoy receptor for RANKL. OPG blocks the association of RANKL, as well as other ligands, with RANK. Almost all other mediators of osteoclastic bone resorption, however, also signal through RANKL, so results obtained from experiments that are designed to block RANKL activity do not reveal the importance of tumour-specific production of RANKL. Moreover, there might be differential expression of RANKL by tumour cells that have metastasized to bone, compared with the same tumour cells at their primary + site, although this has not yet been convincingly documented. Perhaps this will eventually be clarified in animal models in which the RANK RANKL system is rendered ineffective specifically in osteoclastic cells by the use of tissue-specific promoters, or in which RANKL expression is specifically knocked-out in the tumour stroma. Mechanisms of osteoblastic metastasis There is accumulating data to identify the factors that stimulate bone formation that is associated with metastatic tumours (BOX 3). One of the most well-studied mediators is the ubiquitous growth factor endothelin-1, which stimulates bone formation and osteoblast proliferation in bone organ cultures. Endothelin-1 is increased in the circulation of patients with osteoblastic metastases and prostate cancer 32 and is also expressed by breast cancer cell lines that cause osteoblastic metastases 33. Osteoblast proliferation and bone metastasis have both been shown to be inhibited in vivo by endothelin-a-receptor antagonists 33,34. Several other factors, as described below, have also been proposed to be potential mediators of osteoblastic metastasis that is associated with prostate cancer (FIG. 4). The transforming growth factor-β family. Several members of the transforming growth factor-β (TGF-β) family are powerful in vivo stimulators of new bone formation 35, and are candidate mediators of osteoblastic metastasis. TGF-β2 is expressed at high levels by PC3 human prostate cancer cells, and was originally isolated from the human prostate cancer cell line PC3 (REF. 36). TGF-β2 stimulates the proliferation of osteoblasts in vitro, as well as bone formation in vivo. Both normal, and neoplastic human and rat prostate tissues also express a variety of bone morphogenetic proteins (BMPs) namely, BMP2, BMP3, BMP4 and BMP6 mrna 37. Proteases and their activators. There have been several reports that a mitogen for rat CALVARIAL osteoblastic cells has been purified from the conditioned media of PC3 cells, and that the sequences of the first ten amino acids were identical to that of the serine protease urokinase (upa) 38. Overexpression of upa by rat prostate cancer cells has been shown to induce bone metastases in vivo 39, and an amino-terminal fragment of upa has been shown to have mitogenic activity for osteoblasts 40. The carboxy-terminal proteolytic domain might mediate tumour invasiveness or growth-factor activation. Proteases that activate growth factors such as TGF-β have been shown to have important functions in bone 41,42. For example, the latent TGF-β binding protein, which functions to mask TGF-β activity, contains plasmin-sensitive cleavage sites. It is possible that, by activating plasmin, upa prevents the sequestration of TGF-β. Another proteolytic mechanism that might be important for metastasis of prostate cancer cells involves prostate-specific antigen (PSA) a serine protease that is overproduced by prostate carcinoma cells and is used as a marker of tumour burden. PSA can cleave PTHrP 588 AUGUST 2002 VOLUME 2 helsenet.info

6 Box 3 Bone growth factors and bone remodelling The cellular events that are responsible for bone formation include osteoblast proliferation and differentiation, as well as osteoclast apoptosis. Osteoblast proliferation is driven by mitogenic factors, such as transforming growth factor-β (TGF-β), insulin-like growth factors (IGFs) and fibroblast growth factors (FGFs). Osteoclast apoptosis is induced by TGF-β, but also by drugs such as oestrogen and bisphosphonates. During physiological bone remodelling, the release of TGF-β as a consequence of resorption might be the physiological inducer of osteoclast apoptosis and, therefore, be responsible for limiting osteoclast lifespan and impairing continued resorption. Osteoblast differentiation involves expression of the structural proteins of the bone matrix, such as type I collagen, as well as other bone proteins, including alkaline phosphatase and osteocalcin. The bone morphogenetic proteins (BMPs) act predominantly to stimulate osteoblast differentiation. Bone formation involves a cascade of events that, once triggered, continues until the osteoblasts undergo apoptosis, which might be regulated by ambient growth-factor concentrations and influenced by drugs such as corticosteroids and parathyroid hormone. In cancer patients who have osteoblastic metastases, any of these growth factors that are released by the tumour cells could be expected to ultimately cause osteoblast differentiation and subsequent bone formation. IGF1, FGFs, PDGF Osteoclast apoptosis Osteoblast chemotaxis Osteoblast proliferation BMPs Osteoblast differentiation Matrix production Mineralization at the amino terminus 43,44 and could also potentially activate other growth factors that are produced by prostate carcinomas. It is not known whether it actually promotes tumour growth, but one possibility is that it could activate osteoblast-stimulating factors, such as insulin-like growth factor 1 (IGF1) and TGF-β by cleaving them from their binding proteins, or even by cleaving PTHrP to an anabolic fragment (see below). Growth factors. Prostatic cancer cells express large amounts of both acidic and basic fibroblast growth factors (FGFs) 45,46, which are potential mediators of osteoblast proliferation in patients with prostate cancer 47,48. Both acidic (FGF1) and basic FGF (FGF2) stimulate bone formation in vivo 49,50. Izbicka et al. 51 have shown that a human tumour cell line that produces an extended form of FGF2 activates osteoblasts and causes bone formation in vivo. When the human breast cancer cell line MCF-7 is transfected stably with the ERBB2 (also known as HER2/Neu) proto-oncogene, it causes osteoblastic metastases in mice 13,52. These tumour cells have been shown to produce the B isoform of platelet-derived growth factor (PDGF-BB). Conditioned media from these tumour cells promotes bone formation in bone organ cultures, but media from cells that are stably transfected with antisense oligonucleotides to PDGF- BB do not. This work indicates that PDGF-BB is a potential mediator of the osteoblastic response in some tumour types. The bone microenvironment Why do some cancers have such high avidity for bone or any other specific metastatic site, for that matter? One reason might be that most circulating tumour cells pass through the bone marrow, as a consequence of its vascularity. However, there are other highly vascularized organs to which tumour cells rarely metastasize. It is, therefore, probable that the environment of bone provides a particularly fertile ground for the growth and aggressive behaviour of the tumour cells that reach it. The concept that there is a relationship between the seed (tumour cells) and the soil (metastatic site) that determines a cancer s capacity to grow and thrive was first proposed by Stephen Paget more than 100 years ago 53. In the case of bone metastasis by breast cancer cells, we now understand some of the molecular mechanisms that support this concept. As discussed above, breast cancer cells, when present in the bone microenvironment, overproduce PTHrP, and this leads to osteoclastic bone resorption 24. Consequently, active growth factors are released from bone that cause proliferation of breast cancer cells. This stimulates further production of PTHrP, which, in turn, causes more bone loss 54. Studies have also shown that IGF1, which is released during bone resorption, can also induce tumour-cell proliferation under similar circumstances 55. In this way, a vicious cycle (FIG. 5) is set up between the tumour cells and bone: resorbed bone releases TGF-β and IGF1, thereby stimulating tumour-cell proliferation and further PTHrP release, which, in turn, causes more bone FGFs BMPs PDGF Latent TGF-β Active TGF-β Bone formation upa + Proteases Tumour cell IGF IGFBPs IGF PTHrP Inactive PTHrP fragments Figure 4 Model for osteoblastic bone metastases caused by prostate cancer. The production of factors, such as fibroblast growth factors (FGFs), bone morphogenetic proteins (BMPs), platelet-derived growth factor (PDGF) and transforming growth factor-β (TGF-β), by tumour cells can directly stimulate osteoblast activity and subsequent bone formation. Proteases, such as prostate-specific antigen, and protease activators, such as urokinase (upa), can activate latent TGF-β, release IGFs from inhibitory binding proteins (IGFBPs) and inactivate the osteolytic factor parathyroidhormone-related peptide (PTHrP) to promote bone formation. NATURE REVIEWS CANCER VOLUME 2 AUGUST

7 Osteoblast precursor TGF-β IGF1 Growth Ca 2+ RANKL OPG Osteoblasts P TGF-β SMAD MAPK? PTHrP Ca 2+ pump Tumour cell resorption and subsequent release from the resorbed bone matrix of TGF-β and IGF1. The evidence for this concept comes from a series of studies. First, when human breast cancer cells are inoculated into the left ventricle of nude mice, they cause osteolysis in distant skeletal sites that is abrogated by neutralizing antibodies to PTHrP 24. PTHrP is also produced in greater amounts by breast cancer cells that have metastasized to bone than those that have metastasized to other sites [OK? doesn t seem to make sense!] 19,20.TGF-β increases the production of PTHrP by breast cancer cells 54. Breast cancer cells that are stably transfected with P Ca 2+ Figure 5 The vicious cycle hypothesis of osteolytic metastases. Interactions between tumour cells and osteoclasts cause not only osteoclast activation and subsequent bone destruction, but also aggressive growth and behaviour of the tumour cells. Production of the parathyroid-hormone-related protein (PTHrP) by tumour cells enhances osteoclast activation and osteolysis (see FIG. 4). Osteolysis leads to the release of bone-derived growth factors, including transforming growth factor-β (TGF-β) and insulin-like growth factor 1 (IGF1), and raises extracellular calcium (Ca 2+ ) concentrations. The growth factors bind to receptors on the tumourcell surface and activate autophosphorylation (P) and signalling through pathways that involve SMAD (cytoplasmic mediators of most TGF-β signals) and mitogen-activated protein kinase (MAPK). Extracellular Ca 2+ binds and activates a Ca 2+ pump. Signalling through these pathways promotes tumour-cell proliferation and production of PTHrP. Other cytokines might also be involved, such as interleukin (IL)-6, IL-1, IL-11 and IL-18 (not shown). There are many potential points to target in this cycle, including PTHrP, the growth-factor-receptor interactions, and the TGF-β signal-transduction pathway. Bisphosphonates are on the market at present, whereas osteoprotegerin and PTHrP antibodies are in clinical trials. mutant TGF-β receptors that cannot respond to TGF-β do not produce PTHrP, and have markedly reduced osteolytic lesions following inoculation in nude mice 54. IGF1 also promotes breast cancer cell proliferation 56,57, whereas mutant IGF1 receptors reduce the proliferation of breast cancer cells in bone 54. Therapeutic approaches The concept of the vicious cycle alters the approach to the treatment of bone metastases, because it means that osteolysis inhibitors might also decrease bone tumour burden. This concept has recently been supported by in vivo studies. Bisphosphonates, when administered to mice with osteolytic lesions following inoculation of MDA-MB-231 cells, have decreased bone lesions, as expected, but also a decrease in tumour burden 58. Similarly, mice that are treated with neutralizing antibodies to PTHrP, as well as tumours that express mutant TGF-β receptors, experience a decrease in tumour burden 24,54 [OK?]. There is a rationale for developing clinical inhibitors of the bone-resorption process, as so much data now indicates that osteolysis supports the growth and aggressive behaviour of metastatic cancers. Diel et al. 59 and Powles et al. 60 have provided clinical data to support the idea that osteolysis inhibitors also reduce tumour burden in bone. However, there is residual controversy over whether the same effects can be seen in soft-tissue metastases. Studies by Diel and colleagues indicate that osteolysis inhibitors do slow tumour growth in these sites, whereas a study by Powles et al. refutes this idea 60. One study has reported that bisphosphonates can actually promote soft-tissue metastasis 61. This important issue will probably require much larger studies to be resolved definitively. Preclinical data indicate that the bisphosphonates have no effect on soft-tissue metastases if they are administered after metastases are already established. However, if they are given from the time of tumour-cell inoculation, they might increase soft-tissue metastases, presumably by rendering bone unsuitable as a site for metastatic growth. Although definitive recommendations cannot be made until clinical data are available, this data indicate that, although bisphosphonates can be safely administered once metastases are already present, they should be used prophylactically with great caution. Newer bone-resorption inhibitors (TABLE 1) might be even more effective than bisphosphonates. These Table 1 Approaches to treating bone metastases Therapy Mechanism Stage of clinical development Bisphosphonates Block bone resorption; might block tumour-cell mitosis On the market and stimulate tumour-cell apoptosis; alleviate bone pain Osteoprotegerin Prevents RANKL from binding its receptor and Phase II stimulating osteoclasts RANK-Fc Prevents RANKL from binding its receptor and Phase I stimulating osteoclasts PTHrP antibodies Neutralize PTHrP Phase III Vitamin-D analogues Decrease PTHrP production Phase III PTHrP, parathyroid-hormone-related peptide; RANK, receptor activator of nuclear factor-κb; RANKL, receptor activator of nuclear factor-κb ligand. 590 AUGUST 2002 VOLUME 2

8 include direct inactivators of osteoclast activity. For example, OPG the natural decoy receptor to RANK is an extremely powerful and potent inhibitor of bone resorption 62. RANK-Fc is a hybrid chimeric molecule that acts in an identical way to OPG 63. These agents cause the most profound decreases in osteoclastic bone resorption, and can be expected to cause similar effects on tumour burden as do the bisphosphonates. In the case of osteoblastic metastasis, the early preclinical studies of Guise and colleagues 55 indicate that specific inhibitors of endothelin-1 signalling, namely antagonists of the endothelin-a receptor, have beneficial effects on both the bone lesions and the tumour burden. Future research Osteolysis in other cancers. So far, there has been little detailed examination of the factors that might be responsible for osteolysis in tumours other than breast cancer or myeloma. A series of cytokines, including RANK ligand and macrophage inflammatory protein 1α (MIP1α) have been proposed to promote osteolysis by melanoma cells. However, these are probably not the only cytokines that promote bone destruction that is associated with myeloma. It has been indicated that IL-1, lymphotoxin, IL-6, hepatocyte growth factor and PTHrP might all have a subsidiary role 64. In other cancers, factors such as IL-11, IL-18, TGF-β and even prostaglandins might be produced by tumours or by host cells that are activated at the tumour site and are involved in the pathophysiology of hypercalcaemia. There is data that inhibitors of prostaglandin synthesis might be effective at blocking tumour-induced osteolysis in some cases 65. Will bisphosphonates be useful in treating bone metastases that are associated with tumours other than breast cancer, and will any form of bisphosphonate be satisfactory as a treatment for metastasis? These questions cannot be answered definitively at the present time, but it does seem likely that bisphosphonates will also be effective treatments for bone metastases from other cancer types. This is because the cellular mechanisms that are responsible for osteolysis are fundamentally identical, and should be blocked in a similar manner by similar agents. There have been suggestions that some bisphosphonates induce tumour-cell apoptosis, and that this might be a mechanism for decreasing tumour burden in the metastatic site 66. Evidence that bisphosphonates cause apoptosis comes from in vitro studies that involve relatively high concentrations of the drugs, but further studies are required to determine the direct effects of bisphophonates on cancer cells. Osteolytic and osteoblastic factors. Many of the factors that promote osteoblastic metastasis remain to be identified. At present, most evidence supports a role for endothelin-1 in breast cancer metastasis, but there is less convincing data to support its involvement in prostate cancer metastasis due, in part, to the relative absence of animal models of this disease. Many other factors including PDGF, members of the TGF-β family, growth factors and proteolytic systems are also involved, possibly all in the same tumour, but more work needs to be done to tie these mechanisms together. So far, TGF-β and IGF1 have been identified as the bone-derived factors that are involved in the vicious cycle of bone destruction and tumour growth. Other growth factors that are present in the bone microenvironment and released as a consequence of bone resorption, such as the PDGFs, BMPs, FGFs and possibly even extracellular calcium 67,68, might also be involved. It is, in fact, probable that the end result is a combined effect of several factors that act synergistically. Tumour cells in the bone microenvironment have an altered phenotype, compared to the same cells in the primary site or other soft-tissue sites. The best evidence for this is in the increased expression of PTHrP in bone metastases. There are, however, other genes, including the PTH receptor and the oestrogen receptor, that have altered gene expression. TGF-β receptors might be particularly important, because the presence or abundance of TGF-β receptors might be the main influence on the aggressive behaviour of the tumour cell 55. For example, one explanation for the dormancy that occurs in some tumours might be low expression levels of TGF-β receptors on tumour-cell surfaces during the dormant phase 55. Mechanisms of abnormal coupling. Little is understood of the molecular mechanisms that are responsible for the balanced coupling between bone resorption and bone formation that occurs under normal physiological conditions. Furthermore, nothing is known of the mechanisms that are responsible for the distortion in this process that occurs in metastatic cancer growing in bone. Perhaps clarification of the aberrations in the coupling that is responsible for predominantly osteolytic or osteoblastic lesions will also shed light on the mechanisms that are responsible for normal bone remodelling. Metastasis is the single most catastrophic complication of cancer, and understanding the biology of the process should provide not only greater insights into normal cell behaviour, but also lead to new therapies that are specifically designed to limit or prevent this cause of morbidity and mortality in patients with cancer. Our understanding of the cellular and molecular events is improving significantly, and the possibility of such therapy is becoming more realistic. 1. Wingo, P. A., Tong, T. & Bolden, S. Cancer statistics. Cancer J. Clin. 45, 8 30 (1995). 2. Mantyh, P. W., Clohisy, D. R., Kotzenburg, M. & Hunt, S. P. Molecular mechanisms of cancer pain. Nature Rev. Cancer 2, (2000). 3. Honore, P. et al. Osteoprotegerin blocks bone cancerinduced skeletal destruction, skeletal pain and pain related neurochemical reorganization of the spinal cord. Nature Med. 6, (2000). Evidence that the bone resorption inhibitor osteoprotegerin inhibits not just osteolysis but also bone pain in metastatic bone disease. 4. Boyde, A., Maconnachie, E., Reid, S. A., Delling, G. & Mundy, G. R. Scanning electron microscopy in bone pathology: review of methods, potential and applications. NATURE REVIEWS CANCER VOLUME 2 AUGUST

9 Scan. Electron Microsc. 4, (1986). This paper shows, by scanning electron microscopy, evidence of osteoclastic resorption pits at bone metastatic sites, indicating the importance of osteoclasts as cellular mediators of osteolysis. 5. Stewart, A. F. et al. Quantitative bone histomorphometry in humoral hypercalcemia of malignancy: uncoupling of bone cell activity. J. Clin. Endocrinol. Metab. 55, (1982). Histomorphometric evidence of the abnormalities in osteoblastic and osteolytic activity that occurs in patients with bone metastases. This was shown by observing the impaired osteoblastic response to increased osteoclastic resorption. 6. Charhon, S. A. et al. Histomorphometric analysis of sclerotic bone metastases from prostatic carcinoma special reference to osteomalacia. Cancer 51, (1983). 7. Garnero, P. et al. Markers of bone turnover for the management of patients with bone metastases from prostate cancer. Br. J. Cancer 82, (2000). 8. Koga, H. et al. Use of bone turnover marker, pyridinoline cross-linked carboxyterminal telopeptide of type I collagen (ICTP), in the assessment and monitoring of bone metastasis in prostate cancer. Prostate 39, 1 7 (1999). 9. Koizumi, M. et al. Dissociation of bone formation markers in bone metastasis of prostate cancer. Br. J. Cancer 75, (1997). 10. Percival, R. C. et al. Biochemical and histological evidence that carcinoma of the prostate is associated with increased bone resorption. Eur. J. Surg. Oncol. 13, (1987). One of the first studies to indicate that osteoblastic metastases are accompanied by increased bone resorption in patients with prostate cancer. 11. Yoshida, K. et al. Serum concentration of type I collagen metabolites as a quantitative marker of bone metastases in patients with prostate carcinoma. Cancer 80, (1997). 12. Pelger, R. C., Hamdy, N. A., Zwinderman, A. H., Lycklama a Nijeholt, A. A. & Papapoulos, S. E. Effects of the bisphosphonate olpadronate in patients with carcinoma of the prostate metastatic to the skeleton. Bone 22, (1998). 13. Yi, B. et al. PDGF in breast cancer promotes osteosclerotic bone metastases. J. Bone Min. Res. 15 (Suppl 1) A1035 (2000). 14. Adami, S. Bisphosphonates in prostate carcinoma. Cancer 80, (1997). 15. Liotta, L. A. & Kohn, E. C. The microenvironment of the tumour host interface. Nature 411, (2001). An excellent review that cites evidence for the importance of the stroma in cancer-cell behaviour at primary and metastatic sites. 16. Liotta, L. A & Kohn, E. Cancer invasion and metastases. JAMA 263, (1990). 17. Fidler, I. J. Tumor heterogeneity and the biology of cancer invasion and metastasis. Cancer Res. 38, (1978). 18. Zetter, B. R. The cellular basis of site-specific tumor metastasis. N. Engl. J. Med. 322, (1990). 19. Southby, J. et al. Immunohistochemical localization of parathyroid hormone-related protein in breast cancer. Cancer Res. 50, (1990). 20. Powell, G. J. et al. Localization of parathyroid hormonerelated protein in breast cancer metastases: increased incidence in bone compared with other sites. Cancer Res. 51, (1991). Clinical data that shows evidence for increased PTHrP expression in bone metastatic sites in patients with breast cancer. 21. Bryden, A. A., Hoyland, J. A., Freemont, A. J., Clarke, N. W. & George, N. J. Parathyroid hormone related peptide and receptor expression in paired primary prostate cancer and bone metastases. Br. J. Cancer 86, (2002). 22. Miki, T., Yano, S., Hanibuchi, M. & Sone, S. Bone metastasis model with multiorgan dissemination of human small-cell lung cancer (SBC-5) cells in natural killer cell-depleted SCID mice. Oncol. Res. 12, (2001). 30. Ogata, E. PTHrP, paraneoplastic syndrome and cancer metastasis. JBMM [Au: full journal title?], 19, (Suppl. 1) S1 S2 (2001). 23. Henderson, M. et al. Parathyroid hormone-related protein production by breast cancers, improved survival, and reduced bone metastases. J. Natl Cancer Inst. 93, (2001). 24. Guise, T. A. et al. Evidence for a causal role of parathyroid hormone-related protein in the pathogenesis of human breast cancer-mediated osteolysis. J. Clin. Invest. 98, (1996). Human breast cancer cells, when inoculated into the systemic circulation of nude mice, cause osteolytic lesions that are abrogated by neutralizing antibodies to the tumour peptide PTHrP. 26. Zhang, Q. X., Borg, A., Wolf, D. M., Oesterreich, S. & Fuqua, S. A. An estrogen receptor mutant with strong hormoneindependent activity from a metastatic breast cancer. Cancer Res. 57, (1997). 27. De Larco, J. E. et al. A potential role for IL-8 in the metastatic phenotype of breast carcinoma cells. Am. J. Pathol. 158, (2001). 28. Downey, S. E. et al. Expression of the receptor for parathyroid hormone-related protein in normal and malignant breast tissue. J. Pathol. 183, (1997). 29. Gallwitz, W. E. et al. Low doses of 6-thioguanine and 6 mercaptoguanosine inhibit osteolytic bone disease caused by human breast cancer. J. Bone Min. Res. 15 (Suppl. 1), S200 (2000). 31. Zhang, J. et al. Osteoprotegerin inhibits prostate cancerinduced osteoclastogenesis and prevents prostate tumor growth in the bone. J. Clin. Invest. 107, (2001). 34. Yin, J. J. et al. Osteoblastic bone metastases: tumorproduced endothelin-1 mediates new bone formation via the endothelin a receptor. J. Back Musculoskeletal. Rehabil. 14, (Suppl. 1) F400 (1999). Data confirming the molecular mechanisms that are responsible for the vicious cycle in breast cancer metastasis to bone. The authors showed that this involved TGF [or TGF-β?], which was presumably released from the bone matrix during resorption, and the tumour peptide PTHrP. 32. Nelson, J. B. et al. Identification of endothelin-1 in the pathophysiology of metastatic adenocarcinoma of the prostate. Nature Med. 1, (1995). 33. Yin, J. J. et al. Endothelin A receptor blockade inhibits osteoblastic metastases. JBMR [Au: journal title in full?], 15 (Suppl. 1), 1254 (2000). 35. Marcelli, M. et al. A single nucleotide substitution introduces a premature termination codon into the androgen receptor gene of a patient with receptor-negative androgen resistance. J. Clin. Invest. 85, (1990). 36. Marquardt, H., Lioubin, M. N. & Ikeda, T. Complete amino acid sequence of human transforming growth factor typeβ2. J. Biol. Chem. 262, (1997). 37. Harris, S. E. et al. Effects of transforming growth factor-β on bone nodule formation and expression of bone morphogenic protein 2, osteocalcin, osteopontin, alkaline phosphatase, and type 1 collagen mrna in long-term cultures of fetal rat calvarial osteoblasts. J. Bone Min. Res. 9, (1994). 38. Rabbani, S. A. et al. An amino-terminal fragment of urokinase isolated from a prostate cancer cell line (PC-3) is mitogenic for osteoblast-like cells. Biochem. Biophys. Res. Commun. 173, (1990). 39. Achbarou, A. et al. Urokinase overproduction results in increased skeletal metastasis by prostate cancer cells in vivo. Cancer Res. 54, (1994). 40. Rabbani, S. A. et al. Structural requirements for the growth factor activity of the amino-terminal domain of urokinase. J. Biol. Chem. 267, (1992). 41. Dallas, S. L. et al. Characterization and autoregulation of latent TGF-β complexes in osteoblast-like cell lines: production of a latent complex lacking the latent TGF-βbinding protein (LTBP). J. Biol. Chem. 269, (1994). 42. Dallas, S. L. et al. Dual role for the latent transforming growth actor-β binding protein in storage of latent TGF-β in the extracellular matrix and as a structural matrix protein. J. Cell Biol. 131, (1995). 43. Cramer, S. D., Chen, Z. & Peehl, D. M. Prostate specific antigen cleaves parathyroid hormone-related protein in the PTH-like domain: inactivation of PTHrP stimulated camp accumulation in mouse osteoblasts. J. Urol. 156, (1996). 44. Iwamura, M., Hellman, J., Cockett, A. T., Lilja, H. & Gershagen, S. Alteration of the hormonal bioactivity of parathyroid hormone-related protein (PTHrP) as a result of limited proteolysis by prostate-specific antigen. Urology 48, (1996). 45. Matuo, Y. et al. Heparin binding affinity of rat prostate growth factor in normal and cancerous prostate: partial purification and characterization of rat prostate growth factor in the Dunning tumor. Cancer Res. 47, (1987). 46. Mansson, P. E. et al. HBGF1 gene expression in normal rat prostate and two transplantable rat prostate tumors. Cancer Res. 49, (1989). 47. Canalis, E. et al. Effects of endothelial cell growth factor on bone remodeling in vitro. J. Clin. Invest. 79, (1987). 48. Canalis, E., Centrella, M. & McCarthy, T. Effects of basic fibroblast growth factor on bone formation in vitro. J. Clin. Invest. 81, (1988). 49. Mayahara, H. et al. In vivo stimulation of endosteal bone formation by basic fibroblast growth factor in rats. Growth Factors 9, (1993). 50. Dunstan, C. R. et al. Systemic administration of acidic fibroblast growth factor (FGF-1) prevents bone loss and increases new bone formation in ovariectomized rats. J. Bone Min. Res. 14, (1999). 51. Izbicka, E. et al. Human amniotic tumor which induces new bone formation in vivo produces a growth regulatory activity in vitro for osteoblasts identified as an extended form of basic fibroblast growth factor (bfgf). Cancer Res. 56, (1996). 52. Yi, B., Williams, P. J., Niewolna, M., Wang, Y. & Yoneda, T. Tumor-derived PDGF-BB plays a critical role in osteosclerotic bone metastasis in an animal model of human breast cancer. Cancer Res. 62, (2002). 25. Stewart, A. F. PTHrP (1-36) as a skeletal anabolic agent for the treatment of osteoporosis. Bone 19, (1996). 53. Paget, S. The distribution of secondary growths in cancer of the breast. Lancet 1, (1889). The classical paper describing the seed and soil hypothesis of tumour-cell metastasis. 54. Yin, J. J. et al. TGF-β signaling blockade inhibits PTHrP secretion by breast cancer cells and bone metastases development. J. Clin. Invest. 103, (1999). 55. Hiraga, T., Williams, P. J., Mundy, G. R. & Yoneda, T. The bisphosphonate ibandronate promotes apoptosis in MDA- MB-231 human breast cancer cells in bone metastases. Cancer Res. 61, (2001). 56. Rasmussen, A. A. & Cullen, K. J. Paracrine/autocrine regulation of breast cancer by the insulin-like growth factors. Breast Cancer Res. Treat. 47, (1998). 57. Yu, H. & Rohan, T. Role of the insulin-like growth factor family in cancer development and progression. J. Natl Cancer Inst. 92, (2000). 58. Sasaki, A. et al. Bisphosphonate risedronate reduces metastatic human breast cancer burden in bone in nude mice. Cancer Res. 55, (1995). In a preclinical model of human breast cancer metastasis to bone, bisphosphonates reduce not only osteolysis but also tumour burden. 59. Diel, I. J. et al. Reduction in new metastases in breast cancer with adjuvant clodronate treatment. N. Engl. J. Med. 339, (1998). The authors show, in a controversial study, that the bisphosphonate clodronate reduces tumour burden in both bone and non-osseous metastatic sites in patients with breast cancer. 60. Powles, T. J. Tamoxifen for prevention of breast cancer: report of the National Surgical Adjuvant Breast and Bowel Project P-1 Study. J. Natl Cancer Inst. 91, 730 (1999). 61. Saarto, T., Blomqvist, C., Virkkunen, P. & Elomaa, I. Adjuvant clodronate treatment does not reduce the frequency of skeletal metastases in node-positive breast cancer patients: 5-year results of a randomized controlled trial. J. Clin. Oncol. 19, (2001). 66. Lee, M. V., Fong, E. M., Singer, F. R. & Guenette, R. S. Bisphosphonate treatment inhibits the growth of prostate cancer cells. Cancer Res. 61, (2001). 62. Griepp, P. et al. A single subcutaneous dose of an osteoprotegerin (OPG) construct (AMGN 0007) causes a profound and sustained decrease of bone resorption comparable to standard intravenous bisphosphonate in patients with multiple myeloma. Blood (Suppl. 1), A3227 (1998). 63. Oyajobi, B. O. et al. Therapeutic efficacy of a soluble receptor activator of nuclear factor-κb-igg Fc fusion protein in suppressing bone resorption and hypercalcemia in a model of humoral hypercalcemia of malignancy. Cancer Res. 61, (2001). 64. Kuehl, W. M. & Bergsagel, P. L. Multiple myeloma: evolving genetic events and host interactions. Nature Rev. Cancer 2, (2002). 65. Seyberth, H. W. et al. Prostaglandins as mediators of hypercalcemia associated with certain types of cancer. N. Engl. J. Med. 293, (1975). 66. Sanders, J. L. et al. Extracellular calcium-sensing receptor expression and its potential role in regulating parathyroid hormone-related peptide secretion in human breast cancer cell lines. Endocrinology 141, (2000). 67. Sanders, J. L., Chattopadhyay, N., Kifor, O., Yamaguchi, T. & Brown, E. M. Ca2+-sensing receptor expression and PTHrP secretion in PC-3 human prostate cancer cells. Am. J. Physiol. Endocrinol. Metab. 281, E1267 E1274 (2001). 68. Moseley, J. M. et al. Parathyroid hormone-related protein purified from a human lung cancer cell line. Proc. Natl Acad. Sci. USA 84, (1987). 69. Stewart, A. F., Wu, T., Goumas, D., Burtis, W. J. & Broadus, A. E. N-terminal amino acid sequence of two novel tumorderived adenylate cyclase-stimulating proteins: identification of parathyroid hormone-like and parathyroid hormone-unlike domains. Biochem. Biophys. Res. Commun. 146, (1987). 592 AUGUST 2002 VOLUME 2

10 71. Strewler, G. J. et al. N-terminal amino acid sequence of two novel tumor-derived adenylate cyclase stimulating proteins: idenfication of parathyroid hormone-unlike domains. Biochem. Biophys. Res. Commun. 146, (1987). 72. Yates, A. J. P. et al. Effects of a synthetic peptide of a parathyroid hormone-related protein on calcium homeostasis, renal tubular calcium reabsorption and bone metabolism. J. Clin. Invest. 81, 932 (1988). 73. Mundy, G. R. Calcium Homeostasis: Hypercalcemia and Hypocalcemia 2nd edn (Martin Dunitz, London, 1990). 74. Mundy, G. R., Wilkinson, R. & Heath, D. A. Comparative study of available medical therapy for hypercalcemia of malignancy. Am. J. Med. 74, (1983). 75. Fidler, I. J. Modulation of the organ microenvironment for treatment of cancer metastasis. J. Natl Cancer Inst. 87, (1995). Online links DATABASES The following terms in this article are linked online to: Cancer.gov: bone cancer breast cancer lung cancer myeloma ovarian carcinoma pancreatic carcinoma prostate cancer renal carcinoma LocusLink: alkaline phosphatase BMP2 BMP3 BMP4 BMP6 type I collagen endothelin-1 endothelin-a receptor ERBB2 FGF1 FGF2 hepatocyte growth factor IGF1 IGFBPs IKK IL-1 IL-6 IL-8 IL-11 IL-18 JUN lymphotoxin MAPK MIP1a NF-kB oestrogen receptor OPG osteocalcin PDGF PSA PTHrP RANK RANKL SMAD TGF-a TGF-b TGF-b2 upa Medscape DrugInfo: alendronate pamidronate OMIM: Paget s disease FURTHER INFORMATION Access to this interactive links box is free online. Medline Plus Bone Cancer Site: University of Michigan Comprehensive Cancer Center Bone Metastasis Facts: NATURE REVIEWS CANCER VOLUME 2 AUGUST

11 ONLINE Biog Gregory R. Mundy is Professor of Molecular Medicine and Assistant Dean for Clinical Research at the University of Texas Health Science Center, San Antonio, Texas, as well as Deputy Director of the San Antonio Cancer Institute. He received his M.B. and B.S. degrees from the University of Melbourne and his M.D. degree from the University of Tasmania. He was previously Professor and Head, Division of Endocrinology and Metabolism at the University of Texas Health Science Center ( ). His research interests include drug discovery in osteoporosis and the effects of tumours on the skeleton. Databases Cancer.gov bone cancer sion=prov ider&view id=66e65f74-91d7-4e25-a1fef6ece43cc71b breast cancer sion=provider&viewid=53d97cba-89a2-45d4-b55db7b5ad7dc2dd lung cancer myeloma sion=provider&viewid=28de d4-412f-91d cdcc49 ovarian carcinoma sion=prov ider&view id=98f656ef-e2f fca cdd pancreatic carcinoma sion=provider&viewid=60625b32-15f ba538ec9 prostate cancer sion=prov ider&view id=f4c08184-f6a9-49d e59224ac renal carcinoma sion=provider&viewid=a599ac b578- a5d37308aaf5 LocusLink alkaline phosphatase BMP2 BMP3 BMP4 BMP6 type I collagen 0or%20col1a2%20not%205155&ORG=Hs endothelin-1 endothelin-a receptor ERBB2 FGF1 FGF2 hepatocyte growth factor IGF1 IGFBPs 0not%201490%20not%203491%20not%202274%20not%20n ov%20not%20pappa%20not%20stat1&org=hs IKK t%20similar%20%20not%20c6orf5%20not% %20not %20simpl&ORG=Hs IL-1 %20myd88%20not%20traf6&ORG=Hs IL-6 IL-8

12 ONLINE IL-11 IL-18 JUN lymphotoxin MAPK 0or%20mapk2%20or%20mapk3%20or%20mapk4%20or%20 mapk5%20or%20mapk6%20or%20mapk7%20or%20mapk8 %20or%20mapk9%20or%20mapk10%20or%20mapk11%20 or%20mapk12%20or%20mapk13%20or%20mapk14&org= Hs MIP1α NF-κB 0not%20CARD10%20not%20LOC51580%20not%20TANK %20not%20TNFRSF11A&ORG=Hs oestrogen receptor OPG osteocalcin PDGF or%20pdgfb%20or%20pdgfc%20not%201277%20not% %20not% &ORG=Hs PSA PTHrP or%20pthr2&org=hs RANK RANKL SMAD 0or%20smad2%20or%20smad3%20or%20smad4%20or%20s mad5%20or%20smad6%20or%20smad7%20or%20smad8% 20or%20smda9%20or%20smad10%20not%201634%20not% %20not%204609%20not%206498&ORG=Hs TGF-α TGF-β or%20tgfb2%20or%20tgfb3%20not%20eng%20not%20tgfb1 i1%20not%20tgfbi%20not%20tiaf1%20not%20fmod%20not %20mmp1&ORG=Hs TGF-β2 upa [link OK?] Medscape DrugInfo alendronate hemistry.asp?drugcode=1%2d10124&drugname=alen- DRONATE+SODIUM+ORAL&DrugType=1 pamidronate hemistry.asp?drugcode=a%2d6250&drugname=pamidr ONATE+DISODIUM+INTRAVEN%2E&DrugType=1 OMIM Paget s disease

13 R EPORTS the interstimulus interval, the same grating was presented but remained stationary in order to avoid a sudden luminance change, which is known to create a transient cortical response exhibiting poor orientation tuning. We verified by single-unit recordings that standing gratings did not evoke cortical activity. 13. We do not rule out, however, the possibility that a few distinct states were mixed in the PCS spatial pattern when this procedure was applied, because single cortical neurons are known to be selective for more than one visual attribute of the stimulus (for example, orientation, direction, and spatial frequency). 14. This functional map for orientation preference shown in Fig. 1D is defined as a single condition orientation map. Thus far such maps could not be obtained due to the large noise previously associated with voltagesensitive dye imaging. Only the recent technical advances allowed us to obtain reproducible single condition maps (7). To minimize contributions from the intrinsic signals, the data were filtered above 0.6 Hz and the maps were computed with frames from 100 to 150 ms after stimulus onset, much earlier than the onset of the slow intrinsic signal. Normally, maps are revealed by differential imaging only [T. Bonhoeffer and A. Grinvald, J. Neurosci. 13, 4157 (1993)]. The single condition map was validated by comparing it to the differential orientation maps obtained by using two stimuli of orthogonal orientations (Fig. 1B). In addition, the map was confirmed by imaging based on intrinsic signals. The latter two types of maps can be obtained with a much better signal-so-noise ratio. Therefore, we are confident that the PCS shown here corresponds to the relevant functional architecture. 15. The evident variability in both single-unit and population responses could have various sources, including the impact of ongoing activity on evoked activity (2) and relatively large noise in these single-trial images. The values of the correlation coefficients warrant a discussion. The fact that correlation coefficients never reach high values (above 0.6) could manifest the mixing of several cortical states in one PCS during the averaging, or it could result from the residual noise in the recording of instantaneous activity patterns. The values of the correlation coefficients should be compared with the highest value that can be obtained for two maps derived under the same experimental conditions. The correlation between two maps produced by the same stimuli on two independent sets of trials was We converted the correlation coefficient between the population activity and the neuron s PCS into a predicted instantaneous firing, as will be explained later (Fig. 3). We then generated a Poisson spike train using this predicted instantaneous rate. 17. The average correlation coefficient for all five pairs of spontaneous and evoked activity sessions was 0.73 (range 0.65 to 0.81) from five different cats. 18. We expect that spontaneous activity of a cortical neuron can be predicted by this procedure whenever the neuron has response properties that are common to many other neurons. In other cases where the tuning properties are not that robust, or the neuron is a member of neuronal assemblies possessing different spatial structures, other types of analysis, such as clustering techniques, may be required. 19. In Fig. 2 we used the neuron s PCS to predict the spike trains. In Fig. 1, C and D, we showed that the PCS and the map of the functional architecture are very similar. Therefore, here we performed the same analysis using both the functional map and the PCS. The results were very similar. In Fig. 3 we show the results based on the functional map. 20. C. D. Gilbert, Cereb. Cortex 3, 373 (1993); W. Singer, Int. Rev. Neurobiol. 37, 153 (1994); R. Malach, R. B. H. Tootell, D. Malonek, Cereb. Cortex 4, 151 (1994). 21. A. M. Sillito, J. A. Kemp, J. A. Milson, N. Berardi, Brain Res. 194, 517 (1980); M. Tsodyks and T. Sejnowski, Network 6, 1 (1995); R. Ben-Yishai, D. Hansel, H. Sompolinsky, J. Comput. Neurosci. 4, 57 (1997). 22. A. M. Thomson and J. Deuchars, Trends Neurosci. 17, 119 (1994); M. Tsodyks and H. Markram, Proc. Natl. Acad. Sci. U.S.A. 94, 719 (1997); L. F. Abbott, J. A. Varela, K. Sen, S. B. Nelson, Science 275, 220 (1997). 23. B. W. Connors and M. J. Gutnick, Trends Neurosci. 13, 99 (1990); B. Ahmed, J. C. Anderson, R. J. Douglas, K. A. Martin, D. Whitteridge, Cereb. Cortex 8, 462 (1998); D. Hansel and H. Sompolinsky, in Methods in Neuronal Modeling, C. Koch and I. Segev, Eds. (MIT Press, Cambridge and London, 1998), pp D. J. Felleman and D. C. Van Essen, Cereb. Cortex 1, 1 (1991); P. A. Salin and J. Bullier, Physiol. Rev. 75, 107 (1995); S. Ullman, Cereb. Cortex 5, 1 (1995). 25. D. Amit, Modeling Brain Function (Cambridge Univ. Press, New York, 1989); J. Hertz, Introduction to the Theory of Neural Computation (Addison-Wesley, Redwood City, CA, 1991); P. Churchland and T. Sejnowski, The Computational Brain (MIT Press, Cambridge, MA, 1992). 26. We thank K. Pawelzik for illuminating discussions and M. Abeles, A. Aertsen, D. Sagi, and H. Sompolinsky for useful comments on the manuscript. Supported by grants from the German Israel Foundation, the Joint German Israeli Research Program, the Wolfson Foundation, Minerva, and Ms. Enoch. 29 June 1999; accepted 25 October 1999 Stimulation of Bone Formation in Vitro and in Rodents by Statins G. Mundy, 1 * R. Garrett, 1 S. Harris, 2 J. Chan, 1 D. Chen, 1 G. Rossini, 1 B. Boyce, 3 M. Zhao, 1 G. Gutierrez 1 Osteoporosis and other diseases of bone loss are a major public health problem. Here it is shown that the statins, drugs widely used for lowering serum cholesterol, also enhance new bone formation in vitro and in rodents. This effect was associated with increased expression of the bone morphogenetic protein 2 (BMP-2) gene in bone cells. Lovastatin and simvastatin increased bone formation when injected subcutaneously over the calvaria of mice and increased cancellous bone volume when orally administered to rats. Thus, in appropriate doses, statins may have therapeutic applications for the treatment of osteoporosis. Diseases of bone loss are a major public health problem for women in all Western communities. It is estimated that 30 million Americans are at risk for osteoporosis, the most common of these diseases, and there are probably 100 million people similarly at risk worldwide (1). These numbers are growing as the elderly population increases. Despite recent successes with drugs that inhibit bone resorption, there is a clear need for nontoxic anabolic agents that will substantially increase bone formation in people who have already suffered substantial bone loss. There are no such drugs currently approved for this indication. In a search for agents that enhance osteoblast differentiation and bone formation, we looked for small molecules that activated the promoter of the bone morphogenetic protein 2 (BMP-2) gene. We chose this assay because osteoblast differentiation is enhanced by members of the BMP family, including BMP-2 (2), whereas other bone growth factors such as transforming growth factor and the fibroblast growth factors (FGFs) stimulate osteoblast proliferation but inhibit 1 OsteoScreen, 2040 Babcock Road, San Antonio, TX 78229, USA. 2 Department of Medicine, 3 Department of Pathology, University of Texas Health Science Center, 7703 Floyd Curl Drive, San Antonio, TX , USA. *To whom correspondence should be addressed. E- mail: mundy@uthscsa.edu osteoblast differentiation (3). To test the effects of compounds on BMP-2 gene expression, we used the firefly luciferase reporter gene driven by the mouse BMP-2 promoter ( 2736/ 114 base pairs). The gene was transfected into an immortalized murine osteoblast cell line, which was derived from a transgenic mouse in which simian virus 40 (SV40) large T antigen was directed to cells in the osteoblast lineage (4), and the effects on the promoter were assessed by luciferase activity in the cell lysates. We examined more than 30,000 compounds from a natural products collection and identified the statin lovastatin as the only natural product in this collection that specifically increased luciferase activity in these cells. The statins are commonly prescribed drugs that inhibit 3-hydroxy-3-methylglutaryl coenzyme A (HMG Co-A) reductase and decrease hepatic cholesterol biosynthesis, thereby reducing serum cholesterol concentrations and lowering the risk of heart attack (5, 6). We also examined the effects of related statins simvastatin, mevastatin, and fluvastatin in this assay. Each of these compounds was maximally effective at 5 M and had no effects at concentrations lower than 1 M. The increase in luciferase activity was blocked by the immediate downstream metabolite of HMG Co-A reductase, mevalonate (7), which suggests that the effects on bone formation were causally linked to inhibition of this enzyme [although mevalonate may DECEMBER 1999 VOL 286 SCIENCE

14 have cellular effects independent of the cholesterol biosynthesis pathway (8)]. Cultured murine (2T3) or human (MG-63) bone cells exposed to statins showed enhanced expression of BMP-2 mrna, as assessed by Northern (RNA) blot analysis (Fig. 1). This effect appeared to be specific, because the statins did not alter expression from the BMP-4 promoter (Fig. 1) or from promoters derived from the gene encoding interleukin-6, the Fig. 1. Northern blot analyses of effects of simvastatin on human MG-63 osteoblasts. BMP- 2 mrna expression, but not BMP-4 mrna expression, is enhanced by simvastatin. Control media (lane 1) or simvastatin was added to MG-63 osteoblasts to a final concentration of 2 M (lane 2) and 5 M (lane 3), and the cells were then cultured for 48 hours. Blots were developed simultaneously with human BMP-2, BMP-4, and glucose-6 phosphate dehydrogenase (GAPDH). The ratio of BMP-2 to GAPDH after culture with simvastatin at 2 M and 5 M concentrations was 1.5 and 2.8, respectively. The ratio of BMP-4 to GAPDH after culture with simvastatin at 2 M and5 M concentrations was 0.9 and 1.1, respectively. R EPORTS gene encoding parathyroid hormone (PTH) related peptide, or from SV40 and cytomegalovirus. A sandwich enzyme-linked immunosorbent assay for BMP-2 revealed increased protein production in MG-63 cells incubated with simvastatin (a 2.7-fold increase with 2.5 M simvastatin). To investigate the biological effects of statins on bone, we added them to neonatal murine calvarial (skullcap) bones in organ culture (9, 10). Calvaria from 4-day-old pups of Swiss white mice (Harlan Sprague-Dawley) were explanted, dissected free of adjacent connective tissue, placed in tissue culture medium containing 0.1% bovine serum albumin (BSA), and incubated with test compounds for 3 to 7 days. The amount of new bone formation was then assessed morphologically as described in (10) and Table 1. Lovastatin, simvastatin, fluvastatin, and mevastatin each increased new bone formation by approximately two- to threefold, an increase comparable to that seen in this assay after treatment with BMP-2 and fibroblast growth factor 1 (FGF-1). There was also a striking increase in new bone and in osteoblast cell numbers at all stages of differentiation (Fig. 2, A and B, and Table 1). We next injected lovastatin and simvastatin into the subcutaneous tissue overlying the murine calvaria in vivo (11 14). The bone cells of the calvaria are responsive to both bone-resorbing factors and osteoblast-stimulating factors (10 15). This technique requires reproducible placement of small volumes of factors or compounds adjacent to bone. It is minimally invasive, and the calvarial periosteum is not scraped or damaged. Male Swiss ICR (Institute for Cancer Research) white mice, 4 to 5 weeks of age, were injected three times per day for 5 days over the right side of the calvaria with either vehicle or the test compound. Each injection contained the compound dissolved in 50 lof phosphate-buffered saline (PBS) with 2% dimethylsulfoxide and 0.1% BSA. Mice were killed on day 21, and calvaria were removed for histomorphometric analysis. We observed an almost 50% increase in new bone formation after only 5 days of treatment, again comparable to that seen with FGF-1 (Fig. 3, A and B) and BMP-2. However, FGF-1 also increases proliferation of cells in the overlying subcutaneous tissue, an effect not seen with BMP-2 or the statins (Fig. 3). To determine whether the statins stimulate new bone formation when administered systemically, we tested their effect on trabecular Fig. 2. (A) Effects of different statins (all at 1 M) on cultures of murine calvarial bones during 72 hours of culture. The statins have obvious effects in increasing the formation of new bone and enhancing the accumulation of mature osteoblasts. Panel 1 represents bones treated with control media and panels 2 to 5 represent simvastatin, lovastatin, mevastatin, and fluvastatin, respectively. Similar effects are seen with rhfgf-1 and recombinant human BMP-2 (rhbmp-2) (40 ng/ml) (panel 6) and (100 ng/ml) (panel 7), which are positive controls in this experiment. (B) Effects of simvastatin (1 M) added to cultures of explanted murine calvaria for 4 or 7 days. Panels 1 and 2 represent bones treated with control media or simvastatin, respectively, for 4 days. There is marked cellular proliferation and accumulation of mature osteoblasts adjacent to new bone in the bones treated with simvastatin. Panels 3 and 4 represent bones treated with control media or simvastatin, respectively, for 7 days. By this time, there is a more marked increase in the thickness of new bone. SCIENCE VOL DECEMBER

15 R EPORTS Fig. 3. (A) Effect of simvastatin on new bone formation after local subcutaneous injection over the murine calvaria. Mice were injected daily with vehicle alone (panel 1) or with simvastatin at 5 mg/kg/day (panel 2) or at 10 mg/ kg/day (panel 3) for 5 days, and bones were then histologically examined at day 21. There is a marked increase in new bone width in statin-treated versus vehicle-treated mice. Panel 4 represents rhfgf-1 (12.5 g/kg/day) and panel 5 represents rh- BMP-2 (30 g/kg/day) as positive controls. They cause effects similar to those of simvastatin. The arrows indicate the new bone that has been formed as a consequence of simvastatin treatment. (B) Quantitative effects of simvastatin on the width of calvarial bones after local injection into the scalp in (A). Total bone area of the right calvaria (the injected side) was measured from the site of muscle implantation to the suture. The width of the new calvarial bone formed was measured at four points (at 0.5-mm intervals), starting at the muscle implantation site. The mean was calculated for each calvaria measured. The effects are compared with those of FGF-1. Table 1. Effects of simvastatin (experiment 1) and lovastatin (experiment 2) on neonatal murine calvarial bone organ cultures. Explanted bones were cultured for 72 hours, and four bones were in each treatment group. Values are means SEM. Treatment Simvastatin New bone area (per mm ) bone volume in female rats after oral administration. Because osteoporosis occurs most frequently in postmenopausal women, we administered statins to ovariectomized rats as well as rats with intact ovaries (Table 2). At the completion of the experiment, the rats were killed by anesthetic overdose, and the right tibia, the right femur, and the lumbar vertebrae were removed and fixed in formalin. The proximal end of the tibia and the lumbar vertebrae were embedded in paraffin Cells (per 0.3 mm of bone) Lovastatin New bone area (per mm ) Cells (per 0.3 mm of bone) Control BMP-2 (40 ng/ml) * 145 5* * 110 3* Statin M M * 135 4* M * * * 102 5* 0.5 M * * * 132 7* *Significantly greater than control (P 0.01). or plastic, and sections were prepared for histomorphometric analysis. Bone formation rates (BFRs) were measured in paraffin-embedded sections. Before being killed, all animals were treated with a fluorochrome labeling regimen that resulted in the deposition of double-fluorochrome labels on active boneforming surfaces. This regimen consisted of tetracycline [15 mg per kilogram of body weight (15 mg/kg) in 200 l of distilled water] and calcein green [20 mg/kg in 200 l of PBS (ph 7.2)]. These were administered by intraperitoneal injection before killing on days 14 and 4, respectively. We measured bone volume, osteoid volume (in plastic-embedded sections), osteoblast surface, osteoclast surface, and osteoclast number (16). Recombinant human FGF-1 (rhfgf-1) (experiment 1) and synthetic human PTH (amino acids 1 through 34 from the NH 2 - terminal end) (experiment 3) were used as positive controls. The statins caused increases in trabecular bone volume of between 39 and 94% after treatment. This was clearly due to an anabolic (bone-forming) effect because there was a parallel increase in BFRs with the use of dynamic parameters (Table 2). There was a concomitant decrease in osteoclast numbers where these were also assessed. Recently, it has been shown that certain bisphosphonates, which are inhibitors of bone resorption and are widely used as therapy for osteoporosis, also act on the cholesterol biosynthesis pathway (17 19) by targeting enzymes more distal in the mevalonic acid pathway than HMG Co-A reductase. It has been postulated that these enzymes are required for prenylation of small proteins such as Rho and Ras and that interference DECEMBER 1999 VOL 286 SCIENCE