“The greatly compressed bodies of the vertebrae … were so soft they could easily be cut with a knife.”—Harvey Cushing, 1932
THE ADVERSE EFFECTS OF HYPERCORTISOLISM on bone were recognized more than half a century ago,1 but today, the iatrogenic form of the disease has become far more common than Cushing's syndrome and glucocorticoid excess is the third leading cause of osteoporosis following loss of sex steroids and old age. It is estimated that as many as 50% of patients requiring glucocorticoids for the control of pulmonary, rheumatologic, autoimmune, hematopoietic, gastrointestinal disease, or to prevent transplant rejection will ultimately suffer fractures.2 The underlying cause of the fractures in glucocorticoid-induced osteoporosis, as in other forms of osteoporosis, is loss of bone. With glucocorticoid treatment, the loss of bone is biphasic with a rapid initial phase of approximately 12% during the first few months, followed by a slower phase of about 2–5% annually. Both cortical and cancellous bone are lost, but the adverse effects of steroids have a predilection for the axial skeleton. Hence, spontaneous fractures of the vertebrae or ribs are often presenting manifestations of the disorder. Besides these fractures, a frequent accompaniment of long-term glucocorticoid therapy is osteonecrosis—otherwise known as aseptic or avascular necrosis—which causes collapse of the femoral head in as many as 25% of patients.3
Because of the heterogeneity of the underlying conditions—some of which (e.g., rheumatoid arthritis, lymphoma, myeloma, Crohn's disease) independently contribute to skeletal deterioration—wide variations of dose and duration of treatment and lack of a faithful animal model, progress toward the elucidation of the mechanism(s) responsible for the adverse impact of glucocorticoids on the skeleton has been slow. As a result, the management of this condition has remained largely empirical. Recent advances in bone biology, in general, and elucidation of key mechanisms in glucocorticoid-induced osteoporosis, in particular, provide for the first time a convincing explanation of the pathogenesis of the disease and raise hope that more rational therapy may be forthcoming. The purpose of this editorial is to highlight these new developments and to point out their pharmacotherapeutic implications. To appreciate them better, however, one must first understand some basic principles of bone homeostasis.
PHYSIOLOGIC BONE REGENERATION
Basic multicellular unit: The instrument of bone remodeling
During life, the skeleton undergoes remodeling, a periodic replacement of old bone by new, which is responsible for the complete regeneration of the adult skeleton every 10 years. Remodeling is carried out by a team of juxtaposed osteoclasts (in the front) and osteoblasts (in the rear), comprising the basic multicellular unit (BMU). In cortical bone the BMUs tunnel through the tissue, while in cancellous bone they move across the trabecular surface forming a trench. In both situations the cellular components of the BMU maintain a well-orchestrated spatial and temporal relationship with each other. Osteoclasts attach to bone and subsequently remove it by acidification and proteolytic digestion. As the BMU advances, osteoclasts leave the resorption site and osteoblasts move in to cover the excavated area and begin the process of new bone formation by secreting osteoid, which is eventually mineralized. In healthy human adults, 3–4 million BMUs are initiated per year and about 1 million are operating at any moment.4 Even though millions of small packets of bone are constantly remodeled, bone mass is preserved thanks to a remarkably tight balance between the amount of bone resorbed and formed during each cycle of remodeling (Fig. 1, upper panel).
Birth of osteoblasts and osteoclasts
Both osteoblasts and osteoclasts are derived from precursors originating in the bone marrow. The precursors of osteoblasts are multipotent mesenchymal stem cells, which also give rise to bone marrow stromal cells, chondrocytes, muscle cells, and adipocytes, while the precursors of osteoclasts are hematopoietic cells of the monocyte/macrophage lineage. The development of both osteoblasts and osteoclasts is controlled by growth factors and cytokines that are produced in the bone marrow microenvironment and is modulated by systemic hormones and probably by mechanical signals.5 Mesenchymal cell differentiation toward the osteoblast phenotype and osteoclastogenesis are inseparably linked because both are stimulated by the same factors, proceed simultaneously, and the former event is a prerequisite for the latter. It now seems quite clear that in the postnatal bone marrow, commitment of pluripotent mesenchymal precursors to the osteoblastic lineage is initiated by bone morphogenetic proteins (BMPs), the same proteins that are responsible for skeletal development during embryonic life and fracture healing.6 BMPs stimulate the transcription of the gene encoding Cbfa1/Osf2, an osteoblast transcription factor.7 In turn, Cbfa1/Osf2 activates osteoblast-specific genes such as osteopontin, bone sialoprotein, type I collagen, and osteocalcin. Lack of Cbfa1 prevents osteoblast development and remarkably also leads to a paucity of osteoclasts. The mechanistic basis of this phenomenon and the dependency of osteoclastogenesis on mesenchymal cell differentiation has recently been established by the discovery of a membrane-bound cytokine-like molecule, RANK ligand/osteoprotogerin ligand/TRANCE, which is expressed in committed preosteoblastic cells.8 The RANK ligand binds to a specific receptor (RANK) which is expressed in the hematopoietic osteoclast progenitors. This interaction is essential and, together with macrophage colony-stimulating factor (M-CSF), sufficient for osteoclastogenesis. Strong support for the above scenario has been provided by studies from the authors' group demonstrating that BMP-2 and BMP-4 and their receptors are expressed in the postnatal marrow and that they are indeed required for both osteoblast as well as osteoclast development.6 Moreover, the promoter of the RANK ligand gene contains two functional Cbfa1-binding sites.9 Hence, a BMP → Cbfa1 → RANK ligand gene expression cascade in cells of the bone marrow stromal/osteoblastic lineage may well constitute the molecular basis of the linkage between osteoblastogenesis and osteoclastogenesis, with BMPs providing the tonic baseline control of both processes, and thereby the rate of bone remodeling upon which other inputs (e.g. biomechanical, hormonal, etc.) operate (Fig. 2).
Death of osteoclasts and osteoblasts by apoptosis
The BMU has an average life span of 6 months. The average life span of its executive cells, however, is much shorter. The average life span of a human osteoclast is 2 weeks and the average life span of an osteoblast is 3 months. Evidence accumulated during the last couple of years indicates that both osteoclasts and osteoblasts undergo apoptosis. Indeed, after osteoclasts have eroded to a particular distance, either from the central axis in cortical bone or to a particular depth from the surface in cancellous bone, they die by apoptosis and are quickly removed by phagocytes.10 The majority of osteoblasts (50–70%) also die by apoptosis once they have completed their bone-forming tasks.11 The remaining osteoblasts have one of two alternative fates: they can become elongated “lining cells” that cover the newly formed bone surface; or they can be entrapped in the mineralized matrix to become osteocytes, cells characterized by a striking stellate morphology, reminiscent of the dendritic network of the nervous system. Osteocytes represent the most common (∼90%) cell type in bone and are believed to be the sensors of the local need for bone augmentatioin or reduction during functional adaptation of the skeleton, the detection of microdamage, and the transmission of signals that lead to bone repair by remodeling.12
From this brief update of normal bone biology, it becomes clear that the rate of supply of new osteoblasts and osteoclasts and the timing of the death of these cells by apoptosis are critical determinants of the initiation of new BMUs and/or extension or shortening of the lifetime of existing ones. Strong support for this concept has been provided by the elucidation of the pathogenesis of postmenopausal and senile osteoporosis and, specifically, the realization that the bone loss underlying either form of the disease is due to changes in the birth13,14 as well as death rate of bone cells.10 As it will be discussed below, glucocorticoid-induced osteoporosis can also be explained by changes in the birth and death of bone cells.
INHIBITION OF OSTEOBLASTOGENESIS AND PROMOTION OF APOPTOSIS OF OSTEOBLASTS AND OSTEOCYTES: KEY PATHOGENETIC MECHANISMS OF GLUCOCORTICOSTEROID-INDUCED OSTEOPOROSIS
The cardinal histologic features of glucocorticoid-induced osteoporosis are decreased bone formation rate, decreased wall thickness of trabeculae, a strong indication of decreased work output by osteoblasts, and in situ death of portions of bone. Increased bone resorption, decreased osteoblast proliferation and biosynthetic activity, sex-steroid deficiency, as well as hyperparathyroidism resulting from decreased intestinal calcium absorption and hypercalciuria due to defective vitamin D metabolism, have all been proposed as mechanisms for the loss of bone that ensues with glucocorticoid excess.15 However, none of these mechanisms provides a satisfactory explanation for the cardinal histologic features of the condition. Moreover, the evidence put forward in support of each of these mechanisms has been weak and often conflicting. For example, while increased osteoclast surface has been shown in some histologic studies, others have not confirmed this finding and more recent ones have even shown that the number of osteoclasts is decreased.16 Likewise, elevated levels of parathyroid hormone (PTH) have not been a consistent finding in glucocorticoid-induced osteoporosis—and in fact, reduced levels have been reported in some studies.17,18 It is also very unlikely that the adverse effects of glucocorticoids on bone are mediated by sex-steroid deficiency, as they are readily manifested in both eugonadal and hypogonadal subjects. Finally, the circulating levels of vitamin D metabolites are usually normal in patients with glucocorticoid excess making improbable that a defect in vitamin D metabolism contributes in any significant degree to the development of this condition.19
Recent studies from our group provide evidence that the decreased bone formation and osteonecrosis can be accounted for by a suppressive effect of glucocorticoids on osteoblastogenesis (and as expected from the above discussion on osteoclastogenesis); as well as by promotion of apoptosis of osteoblasts and osteocytes.20 Moreover, our work has demonstrated that the mouse, unlike the rat and other previously examined laboratory animals, is a faithful animal model of the glucocorticoid-induced bone loss in humans. Indeed, mice receiving glucocorticoids for 4 weeks—a period equivalent to approximately 3–4 years in humans—exhibited decreased bone mineral density associated with a decrease in the number of osteoblast, as well as osteoclast, progenitors in the bone marrow and a dramatic reduction in cancellous bone area and trabecular width compared with placebo controls. These changes were associated with a significant reduction in osteoid area and a decrease in the rates of mineral apposition and bone formation. More strikingly, glucocorticoid administration to mice caused a 3-fold increase in the prevalence of osteoblast apoptosis in vertebrae and induced apoptosis in 28% of the osteocytes in metaphyseal cortical bone. Although there was a significant correlation between the severity of the bone loss and the extent of reduction in bone formation, some of the bone loss we observed was due to an early increase in bone resorption as evidenced by an increase in osteoclast perimeter by histomorphometric examination of vertebral cancellous bone after only 7 days of steroid treatment. The same histomorphometric changes seen in mice receiving glucocorticoids for 4 weeks were confirmed in biopsies from patients receiving long-term glucocorticoid therapy. Moreover, as in mice, an increase in osteoblast and osteocyte apoptosis was found in the human biopsies. Compared to osteoblast apoptosis, osteocyte apoptosis was far more prevalent, probably because of the anatomical isolation of osteocytes from scavenger cells. Consistent with these in vivo findings, we have more recently established that glucocorticoids promote osteoblast and osteocyte apoptosis in vitro.21 Decreased production of osteoclasts can explain the reduction in bone turnover with chronic glucocorticoid excess, whereas decreased production and apoptosis of osteoblasts can explain the decline in bone formation and trabecular width. Furthermore, accumulation of apoptotic osteocytes may contribute to osteonecrosis. Hence, as in the postmenopausal and senile form of osteoporosis the fundamental problem in the steroid-induced form of the disease is a change in the number of bone cells (Fig. 1, lower panel).
In a follow-up study, the prevalence of osteocyte apoptosis was examined in whole femoral heads obtained from patients with glucocorticoid-induced osteoporosis who had undergone resection of the femoral head and prosthetic hip replacement because of the disease.22 Control specimens were obtained from patients with osteonecrosis due to other clinical disorders including femoral neck cores from five patients with sickle cell disease, a femoral head removed after traumatic fracture and rupture of ligamentum teres, and three femoral heads taken from patients with alcoholism. Abundant apoptotic osteocytes and cells lining cancellous bone were found in the proximal femoral heads resected from patients with glucocorticoid-induced avascular necrosis, whereas apoptotic bone cells were absent from specimens removed because of traumatic or sickle cell disease and were rare in alcohol-induced femoral necrosis. Furthermore, the apoptotic osteocytes were adjacent to the subchondral fracture crescent, whereas empty osteocytic lacunae, the cardinal sign of bone necrosis, were infrequent. Reduced cancellous bone area, increased marrow adipocytes, and decreased hematopoietic marrow were noted in the specimens of the glucocorticoid receiving patients. Although the function of osteocytes in bone is far from being understood, it is believed that these cells may participate in the detection of microdamage and the transmission of signals that lead to its repair by remodeling. In view of this, increased apoptosis of osteocytes may account for the so-called “bone necrosis” associated with glucocorticoid excess. In the past, glucocorticoid-induced osteonecrosis has been attributed to fat emboli, compression of the blood vessels of the femoral head by marrow fat or fluid retention, and poorly mending fatigue fractures. These new findings strongly support the contention that glucocorticoid-induced avascular necrosis is a misnomer; the bone is neither avascular nor necrotic, instead showing prominent apoptosis of cancellous lining cells and osteocytes. Glucocorticoid-induced osteocyte apoptosis, a cumulative and unrepairable defect, could uniquely disrupt the proposed mechanosensory role of the osteocyte network and thus promote collapse of the femoral head. At this stage, it is unknown whether changes in the regulation of osteocyte programmed cell death contribute to the bone loss associated with other forms of osteoporosis. However, it is worth noting that osteocyte apoptosis is increased in estrogen-deficient women and that a significant proportion of osteocytes gradually die with age.23,24
The precise mediators of the cellular changes caused by glucocorticoid excess remain a matter of conjecture. Nonetheless, there is evidence that glucocorticoids decrease the expression of the TGF-β type I receptor secondary to direct suppression of Cbfa125 and they antagonize the effects of BMP2 and insulin-like growth factor I.26,27 Further, the increased adipogenesis noted in the bone marrow of mice and humans with glucocorticoid excess might be due to increased expression of PPARγ2,28 a transcription factor that induces terminal adipocyte differentiation while suppressing osteoblast differentiation.29 In addition, the proapoptotic effect of glucocorticoids on osteoblasts can be prevented by overexpression of the Bcl-2 gene, suggesting that suppression of Bcl-2 or the ratio of Bcl-2 over BAX may also be key mechanisms21 (Table 1).
PAST, PRESENT, AND FUTURE OF THE TREATMENT OF GLUCOCORTICOID-INDUCED OSTEOPOROSIS
Calcium, vitamin D and its metabolites, thiazide diuretics, fluoride, estrogen, testosterone, medroxyprogesterone, nandrolone decanoate, growth hormone, deflazacort—an oxazoline derivative of prednisolone—calcitonin, and bisphosphonates have all been proposed as therapy for glucocorticoid-induced osteoporosis. Unfortunately, with the only exception being the bisphosphonates, none of these drugs has been satisfactory and none has proven antifracture efficacy in glucocorticoid-induced osteoporosis. In fact, for most of them, the benefits of some modest increases in bone mineral density (BMD) have been outweighed by serious side effects. A calcium intake of 1500 mg/day is certainly safe if there is no history of renal calculi, but when used alone bone loss is not prevented. Sex-steroid replacement therapy is useful in hypogonadal patients, if not otherwise contraindicated. Calcitonin is seldom potent enough to have a clinically significant impact.
Etidronate has been shown to improve spinal BMD, but must be given intermittently to avoid drug-induced defects in mineralization.30,31 Therapy with intravenous pamidronate or oral alendronate is more attractive because of greater safety, but the gain in BMD is still modest. Nonetheless, a small reduction in new vertebral fractures was recently reported in patients receiving alendronate (overall incidence, 2.3%; compared with the placebo group, 3.7%; relative risk, 0.6%; 95% confidence interval, 0.1–4.4).32 Bisphosphonates are most effective when started early, before the rapid, early bone loss, but their beneficial effects are less impressive in patients with established disease. In vitro studies in the authors' laboratory show that bisphosphonates prevent osteoblast and osteocyte apoptosis induced by glucocorticoid.33 In view of this evidence, it is theoretically possible that at least part of the antifracture efficacy of these drugs may be due to this mechanism.
Considering our current understanding of the pathophysiology of glucocorticoid-induced osteoporosis, it is obvious that the ideal drug for this condition should be an anabolic agent which would increase bone mass. In a recent trial, Lane et al. reported that daily subcutaneous injections of PTH is a safe and effective treatment for corticosteroid-induced osteoporosis.34 Yet, as mentioned above, PTH has previously been thought of as one of the potential pathogenetic factors for the development of this condition. Can the new insights into the pathogenesis of glucocorticoid-induced osteoporosis and the evidence that intermittent PTH administration is an effective treatment be reconciled on a rational mechanistic basis? In direct contrast to the suppressive effect of glucocorticoids on bone formation, intermittent PTH administration increases bone mass in animals and humans, but the mechanism of this effect had remained an enigma for a long time. We have now obtained evidence that prevention of osteoblast and osteocyte apoptosis is the principal mechanism for the anabolic effect of PTH on bone.35 Thus, in mice, PTH increased the life span of mature osteoblasts by preventing apoptosis, an effect readily reproduced in vitro on osteoblasts and osteocytes, rather than by changing the rate of generation of new osteoblasts. Moreover, PTH prevents in vitro apoptosis induced by either etoposide or dexamethasone in primary cultures of calvaria cells, an osteocyte cell line, as well as in murine and human osteoblastic cell lines.36 Therefore, it seems that PTH and perhaps future PTH mimetics and nonpeptide inhibitors of apoptosis pathways in osteoblasts represent for the first time pathophysiology-based, i.e., rational as opposed to empirical, pharmacotherapies for osteoporosis, in particular, those in which osteoblast progenitor formation is suppressed, such as the osteoporoses associated with glucocorticoid excess and senility.
Finally, it has been reported recently that glucocorticoid-induced apoptosis of thymocytes is mediated by a mechanism that requires the dimerization of the glucocorticoid receptor and direct binding of the receptor to glucocorticoid response element.37 In addition, synthetic glucocorticoids have been developed that exhibit anti-inflammatory activity in vivo as potently as classical glucocorticoids, without requiring glucocorticoid receptor–DNA binding and transactivation.38 The demonstration of osteoblast and osteocyte apoptosis in animals and humans with glucocorticoid-induced osteoporosis raises the possibility that these “designer glucocorticoids” might have bone sparing effects.
The authors acknowledge the support of the National Institutes of Health (P01-AG13918 and R01-43003) and the Department of Veterans Affairs for their research; A. Michael Parfitt, Robert L. Jilka, and Teresita Bellido for helpful discussions; and Tonya Smith for the preparation of this manuscript.