Inhibition of Osteoblastogenesis and Promotion of Apoptosis of Osteoblasts and

Osteocytes by Glucocorticoids
Potential Mechanisms of Their Deleterious Effects on Bone
Robert S. Weinstein, Robert L. Jilka, A. Michael Parfitt, and Stavros C. Manolagas
Division of Endocrinology/Metabolism, Center for Osteoporosis and Metabolic Bone Diseases, University of Arkansas for Medical
Sciences, and the McClellan Veterans Affairs Medical Center GRECC, Little Rock, Arkansas 72205

Abstract Bone loss due to glucocorticoid excess is diffuse, affecting

both cortical and cancellous bone, but has a predilection for
Glucocorticoid-induced bone disease is characterized by de- the axial skeleton. Therefore, spontaneous fractures of the ver-
creased bone formation and in situ death of isolated seg- tebrae or ribs are often presenting manifestations of the disor-
ments of bone (osteonecrosis) suggesting that glucocorticoid der (3, 4). A cardinal feature of glucocorticoid-induced os-
excess, the third most common cause of osteoporosis, may teoporosis is decreased bone formation (5). In addition,
affect the birth or death rate of bone cells, thus reducing patients receiving long-term glucocorticoid therapy sometimes
their numbers. To test this hypothesis, we administered develop collapse of the femoral head (osteonecrosis), but the
prednisolone to 7-mo-old mice for 27 d and found decreased mechanism underlying this is uncertain (6). Decreased bone
bone density, serum osteocalcin, and cancellous bone area formation and in situ death of isolated segments of the proxi-
along with trabecular narrowing. These changes were ac- mal femur suggest that glucocorticoid excess may alter the
companied by diminished bone formation and turnover, as birth and death of bone cells. We have reported previously
determined by histomorphometric analysis of tetracycline- that defective osteoblastogenesis is linked to reduced bone for-
labeled vertebrae, and impaired osteoblastogenesis and os- mation and age-related osteopenia in the SAMP6 mouse (7).
teoclastogenesis, as determined by ex vivo bone marrow cell In addition to the relationship between aberrant osteoblast
cultures. In addition, the mice exhibited a threefold increase production and osteoporosis, we have shown recently that a
in osteoblast apoptosis in vertebrae and showed apoptosis in significant proportion of osteoblasts undergoes apoptosis (8),
28% of the osteocytes in metaphyseal cortical bone. As in which raises the possibility that the premature or more fre-
mice, an increase in osteoblast and osteocyte apoptosis was quent occurrence of osteoblast apoptosis could contribute to
documented in patients with glucocorticoid-induced os- incomplete repair of resorption cavities and loss of bone.
teoporosis. Decreased production of osteoclasts explains the To test the hypothesis that glucocorticoid-induced bone
reduction in bone turnover, whereas decreased production disease is due to changes in the birth or death rate of bone
and apoptosis of osteoblasts would account for the decline cells, we used a murine model of glucocorticoid excess as well
in bone formation and trabecular width. Furthermore, accu- as bone biopsy specimens obtained from patients with gluco-
mulation of apoptotic osteocytes may contribute to osteone- corticoid-induced osteoporosis. In this report, we demonstrate
crosis. These findings provide evidence that glucocorticoid- that glucocorticoid administration decreases bone formation
induced bone disease arises from changes in the numbers of rate and bone mineral density (BMD)1 accompanied by defec-
bone cells. (J. Clin. Invest. 1998. 102:274–282.) Key words: tive osteoblastogenesis and osteoclastogenesis in the bone
bone marrow cells • remodeling • bone formation • osteo- marrow and increases apoptosis of mature osteoblasts and os-
clasts • osteoporosis teocytes.

Introduction Methods
The adverse effects of hypercortisolism on bone have been Animals. Male Swiss Webster mice (Charles River Laboratories,
recognized for more than 60 yr (1), but the precise cellular and Stone Ridge, NY) were electronically tagged (Biomedic Data System
molecular basis of these changes has remained elusive. Today, Inc., Maywood, NJ) and kept in plastic cages (3–5 animals per cage)
the iatrogenic form of the disease has become far more com- under standard laboratory conditions with a 12-h dark, 12-h light cy-
mon than Cushing’s syndrome, and glucocorticoid-induced os- cle and a constant temperature of 208C and humidity of 48%. All
mice were fed a standard rodent diet (Agway RMH 3000, Arlington
teoporosis is now third in frequency after postmenopausal and
Heights, IL) containing 22% protein, 5% fat, 5% fiber, 6% ash, 3.5
senile osteoporosis (2).
kcal/g, 1.0 IU vitamin D3/g, 0.97% calcium, and 0.85% phosphorus
with water ad libitum. The animals and food supply were weighed at
1-wk intervals throughout the experiment. Studies were approved by
Address correspondence to Robert S. Weinstein, M.D., Division of the UAMS Division of Laboratory and Animal Medicine.
Endocrinology and Metabolism, Slot 587, University of Arkansas for Glucocorticoid administration: experimental design. BMD deter-
Medical Sciences, 4301 W. Markham St., Little Rock, AR 72205- minations were done at 2-wk intervals to identify the peak adult bone
7199. Phone: 501-686-5130; FAX: 501-686-8148; E-mail: rweinstein@ mass of the mice, which was reached between 5 and 6 mo of age (9).
medlan.uams.edu We used animals at peak bone mass to avoid obscuring the negative
Received for publication 12 January 1998 and accepted in revised
form 1 May 1998.
1. Abbreviations used in this paper: BMD, bone mineral density;
The Journal of Clinical Investigation CFU-F, CFU-fibroblast; CFU-OB, CFU-osteoblast; TRAPase, tar-
Volume 102, Number 2, July 1998, 274–282 trate-resistant acid phosphatase; TUNEL, transferase-mediated bi-
http://www.jci.org otin-dUTP nick end-labeling.

274 Weinstein et al.

impact of glucocorticoid excess on BMD by the confounding effects Livers were examined for fatty infiltration as a sign of prednisolone
of increased linear and radial growth. Before the experiment began, toxicity. The weight of the seminal vesicles (mg/100 g body wt) was
BMD measurements were repeated to allocate the animals into used as an index of the androgen status of the animals (12). To help
groups (n 5 4–5) with equivalent spinal density values. The mice (7 interpret these measurements, a separate group of animals was or-
mo old) received placebo or prednisolone, a synthetic glucocorticoid chidectomized (n 5 5).
analogue that does not require hepatic hydroxylation and has mini- Bone densitometry. Dual-energy x-ray absorptiometry (DEXA)
mal mineralocorticoid activity, thus eliminating the need for potas- was used to determine global (whole body minus the head), spinal,
sium supplementation or sodium restriction (10, 11). It was found and hindquarters BMD in live mice (7, 9). The scans done at 27 d af-
that implantation of pellets releasing 0.5 mg/kg/d of prednisolone (the ter pellet implantation were analyzed using the “Compare” tech-
no effect dose) did not decrease BMD (data not shown). Therefore, nique, in which the evaluation is based on the exact positioning and
we used two doses, 0.7 (lower dose) and 2.1 mg/kg/d (higher dose), region of interest placement of the baseline scan. Accuracy of the
chosen from pilot studies to bracket the dose (1.4 mg/kg/d) that in- DEXA measurements was demonstrated by the strong linear rela-
variably causes loss of bone density. These doses were administered tionship between ash weight and bone mineral content at each region
for 27 d by subcutaneous implantation of slow-release pellets (Inno- (7). Over 18 mo, the coefficient of variation for the BMD of a plastic-
vative Research of America, Sarasota, FL). BMD measurements embedded whole mouse skeleton was 3.0% (n 5 146).
were obtained at the beginning of the experiment and 27 d after im- Serum and urine biochemical measurements. Serum osteocalcin was
plantation. For dynamic histomorphometric measurements, tetracy- measured by radioimmunoassay using a goat anti–murine osteocalcin
cline HCl (30 mg/kg body wt) was given intraperitoneally 17 and 23 d and murine osteocalcin as tracer and standard (Biomedical Technolo-
after implantation. After 27 d, the mice were killed, serum and urine gies, Stoughton, MA). Urinary free deoxypyridinoline excretion was
specimens were taken, bone marrow aspirates were obtained from determined by a microtiter competitive enzyme immunoassay (Pyri-
the right femur for ex vivo marrow cell cultures, and the left femur links-D; Metra Biosystems, Mountain View, CA) and was expressed
and lumbar vertebrae were prepared for histomorphometric analysis. as a ratio to the urinary creatinine.

Figure 1. Photomicrographs of the effects of prednisolone on murine vertebral cancellous bone. (A) Longitudinal, panoramic section from a
mouse receiving placebo and (B) section from a mouse receiving prednisone. The histomorphometric reading area is outlined. Toluidine blue
stain, original magnification 325.

Regulation of the Birth and Death of Osteoblasts by Glucocorticoids 275

 Bone histomorphometric analysis. The distal femora and lumbar Table I. BMD and Serum and Urine Biochemical
vertebrae were fixed in 48C Millonig’s phosphate-buffered 10% for- Measurements in Prednisolone-treated Mice
malin, pH 7.4, embedded undecalcified in methyl methacrylate, and
stained as described previously (7, 9, 13). The histomorphometric ex- Measurement Placebo 0.7 mg/kg/d 2.1 mg/kg/d
amination was done with a computer and digitizer tablet (version
3.00; OsteoMetrics Inc., Atlanta, GA) interfaced to a Zeiss Axio- Global BMD (% change) 22.762.1 25.062.2* 26.661.9‡
scope (Carl Zeiss, Inc., Thornwood, NY) with a drawing tube attach- Spinal BMD (% change) 23.163.0 26.863.2 28.763.5*
ment. All cancellous measurements were two-dimensional, confined Hindquarters BMD
to the secondary spongiosa, and made at a magnification of 400 (nu- (% change) 0.4610.4 23.868.0 23.466.9
merical aperture 0.75) (Fig. 1). The terminology and units used are Osteocalcin (mg/liter) 93.8611.5 63.0627.7* 46.4613.8‡
those recommended by the Histomorphometry Nomenclature Com- Deoxypyridinoline
mittee of the American Society for Bone and Mineral Research (14).
(mM/mM creatinine) 78.369.3 63.6614.7 81.5611.3
The trabecular width and osteoid width were measured directly. Tra-
becular spacing and number were calculated (15). Only tartrate-resis-
tant acid phosphatase (TRAPase)-positive cells were included in the Data shown are the mean6SD from five to seven animals. * P , 0.05 vs.
osteoclast perimeter. The rate of bone formation (mm2/mm/d) and placebo; ‡ P , 0.005 vs. placebo.
turnover (%/d) were calculated as described previously (7).
Detection and quantification of osteoblasts and osteoclasts in ex
vivo bone marrow cultures. One femur from each mouse was flushed
with 5 ml of phenol red–free aMEM (GIBCO BRL, Gaithersburg, then incubated with 0.5% pepsin for 30 min at 378C. Apoptotic cells
MD) containing 10% FBS (Hyclone, Logan, UT) to obtain marrow were detected by the TUNEL reaction (transferase-mediated biotin-
cells. After the cells were rinsed and resuspended to obtain a single dUTP nick end-labeling) using Klenow terminal deoxynucleotidyl
cell suspension, the nucleated cell count was determined using a transferase (Oncogene Research Products, Cambridge, MA) in sec-
Coulter Counter (Coulter Corp., Miami, FL). Cells from each animal tions counterstained with 1% methyl green. The TUNEL reaction
were cultured separately. was noted within cell nuclei and the cells whose nuclei were clearly
The number of CFU-fibroblast (CFU-F) and CFU-osteoblast brown from the peroxidase-labeled antidigoxigenin antibody instead
(CFU-OB) present in the bone marrow preparations was determined of the blue-green from the methyl green were interpreted as positive.
as described previously (16–18). Briefly, cells were seeded at 1.5 3 Plastic-embedded sections of weaned rat mammary tissue were used
106 per 10 cm2 well for the determination of CFU-F number and as a positive control. Negative controls were made by omitting the
maintained for 10 d in phenol red–free aMEM containing 15% prese- transferase. Morphological changes characteristic of apoptosis were
lected FBS, 50 mM ascorbic acid, and 10 mM b-glycerophosphate examined carefully to minimize ambiguity regarding the interpreta-
(Sigma Chemical Co., St. Louis, MO) with one-half of the medium re- tion of results. With these precautions, TUNEL has been unequivo-
placed after 5 d. After fixation in neutral buffered formalin and stain- cally associated with apoptosis (19). In addition, TUNEL has been
ing with hematoxylin, colonies containing a minimum of 20 fibroblas- used with DNA fragmentation and immunohistochemical studies to
toid cells were enumerated. Cells were seeded at 2.5 3 106 cells per demonstrate apoptosis of osteoblastic cells and osteoblasts both in
10-cm2 well for the determination of CFU-OB number and cultured vitro and in vivo (8, 20). Apoptosis was also assessed in transiliac
for 25–28 d as described above for CFU-F. After fixation in 50% eth- bone biopsy specimens taken from 2 patients with glucocorticoid-in-
anol and 18% formaldehyde, cultures were stained using Von Kossa’s duced osteoporosis (22 and 36 yr old, receiving 15–25 mg/d of pred-
method to visualize and enumerate colonies containing mineralized nisone for 3–6 yr) and from 12 age-, sex-, and race-matched controls
bone matrix. (13). Two longitudinal sections were examined from each patient and
Osteoclast formation in bone marrow cultures was assessed in control subject. Osteoblasts were identified as cuboidal cells lining
replicate cultures (4–6 from each animal) maintained for 9 d in the the osteoid-covered trabecular perimeter (7, 9, 13). Osteocytes were
presence of aMEM, 10% FBS, and 10 nM 1,25(OH)2D3 as described identified inside lacunae in mineralized bone.
previously (7). In brief, marrow cells were cultured at 1.5 3 106 per Statistics. We tested for differences in the bone densitometry val-
2-cm2 well on 13-mm round Thermanox disks and maintained for 8 d ues using the percent change in BMD from baseline. Dose–response
in the presence of 10% FBS in aMEM supplemented with 1028 M relations were tested by one-way ANOVA. To further evaluate
1,25(OH)2D3 (provided by Dr. Milan Uskokovic, Hoffman-LaRoche, changes in bone histomorphometry, we also used Student’s t test to
Nutley, NJ). At the end of the experiment, cells were processed for assess for significant differences between group means, after testing
the autoradiographic detection of bound 125I-calcitonin (125I-CT) and for equivalence of variances and normal distribution of data. The sig-
stained for TRAPase. Because many osteoclasts in murine bone pos- nificance of the relative frequency of apoptotic cells was determined
sess only one nucleus (7), it is impossible to distinguish between pre- with the x2 statistic. P , 0.05 was considered significant (21).
osteoclasts and mononuclear osteoclasts in ex vivo cultures of murine
bone marrow cells. Therefore, mononucleated and multinucleated
cells that both bind 125I-CT and express TRAPase were designated as
osteoclastic cells. The number of osteoclasts formed in this assay is a
Table II. Food Intake, Body Weight, and Seminal Vesicle
reflection of the number of osteoclast progenitors present in the bone
Weight in Prednisolone-treated Mice
marrow aspirate and the number of stromal/osteoblastic support cells
that form during the culture period. Measurement Placebo 0.7 mg/kg/d 2.1 mg/kg/d
The number of CFU-F colonies, CFU-OB colonies, and osteo-
clastic cells formed from the marrow cells of each animal was ex- Food intake (g/d) 3.460.6 3.660.2 3.760.4
pressed as the number per femur, which was calculated by multiply- Body weight (g) 37.966.0 33.864.3 32.264.2
ing the number of colonies or osteoclasts obtained per 106 cells Seminal vesicle weight
seeded at the initiation of the cultures by the total number of marrow
(mg/100 g body weight) 74.6614.6 92.768.7 83.166.9
cells obtained from the animal.
Measurement of apoptosis in undecalcified bone sections. Sections
were mounted on silane-coated glass slides (Scientific Device Lab, Data shown are the mean6SD. Seminal vesicle weight in a separate or-
Inc., Des Plains, IL), deplasticized, and incubated in 10 mM citrate chidectomized control group was 11.363.1 mg/100 g body wt, P , 0.001
buffer, pH 7.6, in a microwave oven at 988C for 5 min. Slides were vs. treated mice.

276 Weinstein et al.

Table III. Vertebral Cancellous Bone Histomorphometry in Swiss Webster Mice after 27 d of Prednisolone Administration

Histomorphometric determination Placebo 0.7 mg/kg/d 2.1 mg/kg/d

Bone area/tissue area (%) 10.461.4 6.962.1 6.361.7‡

Trabecular width (mm) 48.062.4 48.664.3 37.164.4‡
Trabecular spacing (mm) 423669 7126302 5466125
Trabecular number (per mm) 1.6660.66 1.4460.47 1.7760.33
Osteoid area/bone area (%) 2.160.2 2.260.8 1.560.2‡
Osteoid perimeter/bone perimeter (%) 15.162.1 15.865.1 9.961.1‡
Osteoid width (mm) 2.660.4 2.060.3 1.960.3*
Osteoblast perimeter/bone perimeter (%) 1.260.9 2.260.2 0.560.4
Osteoclast perimeter/bone perimeter (%) 2.761.1 2.660.5 1.161.7
Reversal perimeter/bone perimeter (%) 2.562.3 3.262.2 7.261.1‡
Mineralizing perimeter/bone perimeter (%) 12.960.5 13.965.6 9.562.5*
Mineral appositional rate (mm/d) 1.2360.11 0.9660.11* 0.7460.20‡
Bone formation rate/bone perimeter (mm2/mm/d) 0.1560.02 0.1360.04 0.0760.03‡
Bone turnover (%/d) 0.6860.09 0.4660.12* 0.2460.11‡

Data shown are the mean6SD. There are four to five animals per group. *P , 0.05 vs. placebo; ‡ P , 0.01 vs. placebo.

Results spacing and, in the lower dose group, decreased trabecular

number, indicating that some trabecular profiles were entirely
Demonstration of bone loss in mice receiving prednisolone. In resorbed. In the higher dose group, osteoid area decreased by
mice implanted with the higher dose of prednisolone, global 29%, osteoid perimeter by 34%, and osteoid width by 27%
and spinal BMD at 27 d were significantly lower than those (P , 0.01). A trend toward decreased osteoblast and osteo-
found in the mice that were implanted with placebo pellets clast perimeters was found in the animals receiving the higher
(Table I). The decrease in global BMD was dose dependent dose. However, there was a threefold increase in the empty
(P , 0.05). Demonstrating the expected propensity for the ax- erosion cavities (devoid of osteoclasts) or reversal perimeter.
ial skeleton, glucocorticoid-induced loss of BMD was less con- The tetracycline-based histomorphometry showed that pred-
spicuous at the hindquarters. The level of serum osteocalcin, nisolone administration caused a 26% decrease in the mineral-
a marker of osteoblast activity, was decreased . 50% when izing perimeter (P , 0.05). In addition, a dose-dependent de-
compared with placebo, whereas urinary deoxypyridinoline crease in the mineral appositional rate was noted (P , 0.05);
excretion was not significantly different between the groups this decline was 22% with the lower dose and 40% with the
(Table I). These effects were not due to changes in food in- higher dose. Furthermore, there was a 53% decrease in the
take, body weight, or androgen status (Table II). In addition, rate of bone formation with the higher dose (P , 0.01), which
hepatic fatty infiltration was absent. correlated with the vertebral cancellous bone area (r 5 0.57,
Effects of glucocorticoid administration on vertebral bone P , 0.05), indicating that the glucocorticoid-induced decreases
histomorphometry. Consistent with the BMD results, in the in bone area were associated with a reduction in the rate of
animals receiving the higher dose there was a 40% decline in bone formation. Bone turnover, expressed as a percentage of
the vertebral cancellous bone area (Fig. 1) and a 23% decline the bone area per day, also decreased in a dose-dependent
in trabecular width (P , 0.01) (Table III). In both predniso- manner (P , 0.05).
lone groups, there was a trend toward increased trabecular Effects of glucocorticoid administration on osteoblastogen-

Figure 2. Quantification of CFU-OB and

osteoclast progenitors formed in ex vivo
bone marrow cell cultures. Marrow cells
were obtained from the femurs of male
mice after 27 d of exposure to placebo
(white bars) or 2.1 mg/kg/d of predniso-
lone (black bars). Cells from each mouse
were cultured separately as described in
Methods.

Regulation of the Birth and Death of Osteoblasts by Glucocorticoids 277

 Figure 3. Effect of prednisolone
on murine osteoblast apoptosis.
Osteoblasts were counted in un-
decalcified sections of cancellous
bone from the vertebral second-
ary spongiosa. The placebo
group is shown in A and the
higher dose prednisolone group
is shown in B. Apoptotic cells in
this experiment were identified
using TUNEL and morphomet-
ric features such as nuclear frag-
mentation and condensation of
chromatin (arrows). Methyl
green counterstain viewed with
Nomarski differential interfer-
ence microscopy, original magni-
fication 3400.

esis and osteoclastogenesis. In bone marrow cell cultures from (3756257 SD vs. 54614, P , 0.05) and the number of osteo-
the animals receiving the higher dose, there was no significant clastic cells formed in response to 1,25(OH)2D3 in ex vivo mar-
change in CFU-F colonies (1,2506374 vs. 6986104, NS). How- row cultures decreased by 65% (13876920 vs. 4926311, P ,
ever, the number of CFU-OB colonies decreased by 86% 0.05) (Fig. 2).

Figure 4. Effect of prednisolone on murine osteo-

cyte apoptosis. The cells were counted in undecal-
cified sections of femoral metaphyseal cortical
bone. The placebo group is shown in A and the
higher dose prednisolone group is shown in B.
Apoptotic osteocytes (arrowheads) are seen in
close proximity to normal cells. Methyl green
counterstain viewed with Nomarski differential in-
terference microscopy, original magnification
3630.

278 Weinstein et al.

 Effects of glucocorticoid administration on apoptosis. teoblasts and osteocytes were clearly identified in both (Fig.
Counting a total of 973 osteoblasts, there was a threefold in- 5, B and C) but were absent from specimens taken from 12
crease in osteoblast apoptosis in the vertebral cancellous bone age-, sex-, and race-matched controls (Fig. 5 A). As in our mu-
of mice receiving the higher dose of prednisolone when com- rine model, bone histomorphometry from these two patients
pared with controls (2.0360.34 vs. 0.6660.07%, P , 0.05). showed the changes expected with chronic glucocorticoid ther-
Morphological changes typical of apoptosis accompanied the apy (5): reduced cancellous bone area (11.1 and 8.8%, normal
TUNEL-positive osteoblasts and included sharply defined, is 22.461.2 SEM), decreased trabecular width (62 and 118 mm,
condensed chromatin plastered against the nuclear membrane, normal is 16169), decreased osteoblast perimeter (2.1 and
nuclear fragmentation, and cell shrinkage (Fig. 3, A and B). 2.3%, normal is 7.660.4), decreased osteoclast perimeter (0
In addition, prednisolone caused the appearance of apop- and 0.4%, normal is 0.960.2), increased reversal perimeter
totic osteocytes in cortical bone sections taken from femora (13.5 and 15.4%, normal is 6.960.7), and diminished bone for-
(Fig. 4, A and B). Whereas none of the osteocytes exhibited mation rate (0.02 and 0.05 mm2/mm/d, normal is 0.09560.012).
apoptotic features in the control animals, 28% of 131 cortical In the cancellous bone of these specimens, z 5% of the osteo-
osteocytes were apoptotic in the animals receiving the higher cytes and 30% of the osteoblasts were apoptotic. Apoptosis of
dose. Osteocyte apoptosis was restricted to small groups of osteoclasts or cortical osteocytes was not observed. A transil-
cells in the center of the femoral metaphyseal cortex and were iac bone biopsy represents a much smaller sample of the hu-
absent from vertebral cortical bone. The apoptotic osteocytes man skeleton than the murine femur and lumbar vertebrae
were identified in close proximity to normal osteocytes, in con- represent of the mouse skeleton. Therefore, it is not surprising
trast to the large homogenous areas of dead and dying cells that the percentage of apoptotic osteoblasts and osteoclasts
typical of cell necrosis. An increase in apoptotic hypertrophic was different in the human and murine specimens.
chondrocytes and bone marrow cells was also noted in mice re- Early effects on bone resorption. Finally, because of the evi-
ceiving either dose of prednisolone. Osteoclast apoptosis was dence for decreased trabecular number, we investigated the possi-
not observed. bility that glucocorticoids initially accelerate bone resorption in
Demonstration of apoptotic osteoblasts and osteocytes in the mouse. To do this we examined the vertebral cancellous
patients with glucocorticoid-induced osteoporosis. In transiliac bone histology in an additional group of somewhat younger
bone biopsies taken from two patients, TUNEL-positive os- mice (5 mo old) after a 7-d administration of the higher dose of

Figure 5. Effect of chronic prednisone treatment on apoptosis in human bone. TUNEL-positive osteoblasts (arrowheads) and osteocytes (ar-
rows) were absent from normal subjects (A) but were clearly identified in patients with prednisone-induced osteoporosis (B and C). Approxi-
mately 5% of the osteocytes and 30% of the osteoblasts were apoptotic. The photomicrographs are from transiliac bone biopsy specimens.
Methyl green counterstain viewed with Nomarski differential interference microscopy, original magnification 3630.

Regulation of the Birth and Death of Osteoblasts by Glucocorticoids 279

prednisolone or placebo (n 5 5). We found that whereas pred- the osteoclast perimeter, normal urinary deoxypyridinoline ex-
nisolone caused a 59% decrease in the osteoblast perimeter cretion, and profound decrease in osteoclastogenesis. The per-
(5.2%61.5 SD vs. 2.161.1, P , 0.005), the osteoclast perime- sistent increase in erosion cavities devoid of osteoclasts, mea-
ter increased 96% (0.51%60.34 vs. 1.0060.41, P , 0.05). sured as the reversal perimeter, merely indicates delayed bone
formation (38), and has been observed previously in glucocor-
Discussion ticoid-treated patients (5, 39). Consequently, we will empha-
size the relevance of our findings at 27 d to chronic, rather than
Our choice of the mouse for these studies was based on our short-term, glucocorticoid administration to humans.
previous experience with its validity as a model of the bone We have demonstrated previously that vertebral cancellous
loss associated with loss of sex steroids (9, 22) and with senes- bone in adult mice undergoes sequential, coupled bone remod-
cence (7), but we also found the mouse to have several advan- eling that is qualitatively similar to that occurring in human
tages over previously used laboratory animals (Table IV). bone (7, 9). Many of the changes in cellular, osteoid, and tetra-
Only in the mouse does glucocorticoid administration consis- cycline-based histological indices induced by glucocorticoid
tently induce axial greater than appendicular bone loss without administration can be accounted for by a reduction in the acti-
weight loss or hypogonadism, accompanied by histological in- vation frequency of bone remodeling, the main determinant of
dices of impaired osteoblast function, thus reproducing the the rate of bone turnover (40), which is an inevitable conse-
major features of the human disease (2–5). Although the doses quence of the substantial decrease in osteoclastogenesis that
used in our studies were higher in relation to body weight than we observed. Although a reduction in bone turnover will not
in humans, they were only mildly higher than the dose deter- by itself cause bone loss, the decrease in trabecular width,
mined by serial bone densitometry to have no effect and were which was the major structural change observed, is usually the
consistent with the much higher metabolic clearance of gluco- result of incomplete cavity repair. This is, at least in part, due
corticoids and other compounds in small laboratory animals to inadequate osteoblast recruitment, either from diminished
than in humans (35–37). Nonetheless, the similarity of the glu- production or ineffective migration to the bone surface (40).
cocorticoid-induced increases in apoptotic cells and bone his- The reduction in osteoblastogenesis was of sufficient magni-
tomorphometric features in mice and humans clearly indicates tude to explain the decrease in bone formation rate, and would
that our observations in the mouse cannot be due to pharma- also have contributed to the inadequate osteoblast recruitment
cological differences. and consequent decline in trabecular width. Thus, the inhibi-
We examined the effects of glucocorticoids after 27 d, a pe- tory effect of glucocorticoids on early bone cell progenitors in
riod equivalent in the mouse to z 3–4 yr in humans. Thus, our the bone marrow can account for many of the in vivo observa-
findings represent long-term rather than acute effects. Al- tions.
though we found a significant correlation between the severity Our data also bear on recent ideas concerning the relation-
of the cancellous bone loss and the extent of reduction in bone ships between early osteoblast and osteoclast progenitors in
formation, several other lines of evidence imply that some of the bone marrow. Although mature osteoclasts and osteo-
the bone loss we observed was due to an early increase in bone blasts are needed successively at each bone surface site that is
resorption which had subsided by the time of examination. being remodeled, these cells are needed simultaneously as the
First, there was suggestive evidence of complete loss of some basic multicellular unit, which is the instrument of bone re-
trabeculae (Table III). Second, based on the bone turnover modeling, progresses through or across the surface of bone
measured in the placebo group, which must be close to the rate (41). The necessary parallel production of executive cells is ac-
found in all the animals at the beginning of the study, even complished by signals that originate from early members of
with total suppression of bone formation the initial rate of the stromal cell–osteoblast family, which support in various
bone turnover could have accounted only for an exponential ways the production of mononuclear preosteoclasts in the
decline in cancellous bone area of 18%, whereas a 40% de- bone marrow (42). Our demonstration of a marked reduction
crease was observed. Finally, we confirmed an early increase in in the numbers of both CFU-OB and osteoclast progenitors
osteoclast perimeter by histomorphometric examination of derived from ex vivo bone marrow cell cultures makes it likely
vertebral cancellous bone after 7 d of prednisolone administra- that glucocorticoid administration inhibits the proliferation
tion. and/or differentiation of the stromal cell–osteoblast family at
By 27 d of prednisolone administration, bone resorption an early stage, leading to a reduction in the number of mature,
fell to or below normal, as indicated by the downward trend in matrix-secreting osteoblasts as well as the osteoblastic cells

Table IV. Confounding Factors with Glucocorticoid Administration to Other Animals

Animals Factors

Rats (23, 34) Paradoxical increase in cancellous bone mass,* decreased food intake and weight
Rabbits and dogs (25–27) Inconsistent changes in bone density and cancellous bone area, weight loss, hepatic fatty infiltration
Ewes (28–30) Histological changes resemble glucocorticoid-treated patients but corresponding changes in bone density and
cancellous bone area are inconsistent

*Glucocorticoids inhibit bone resorption and promote apoptosis in rat osteoclasts in vitro (31), whereas bone resorption is stimulated in neonatal
mouse calvaria (32). An additional species difference is that glucocorticoids stimulate bone nodule formation from rat calvarial cells in vitro (33) but
inhibit differentiation in a murine osteoblastic cell line (34).

280 Weinstein et al.

that support osteoclast development. A direct inhibitory effect teocyte death in cancellous bone, indicated by absence of lactic
of glucocorticoids on osteoclast precursor proliferation is not dehydrogenase activity, increases in prevalence with age in the
excluded by our data, but would be less easy to reconcile with upper femur but not in the vertebrae (49), probably because of
the finding of an early increase in the osteoclast perimeter. the higher bone turnover in the spine. Empty lacunae and en-
It has long been known that some osteoblasts become os- zyme absence can reveal the fact but not the mode of death.
teocytes and some become lining cells, but these fates com- Osteocyte apoptosis has been detected recently in human iliac
bined do not account for all the osteoblasts initially present. cancellous bone and its prevalence was increased by pharma-
Although migration along or away from the bone surface is cological induction of estrogen deficiency (19). We have now
possible, death has always seemed the most likely alternative demonstrated that chronic glucocorticoid administration, both
fate (43). Recently, we provided evidence for this possibility by to mice and to human patients, likewise increases the preva-
demonstrating that osteoblasts in remodeling bone undergo lence of osteocyte apoptosis. The proportion of apoptotic os-
apoptosis with a frequency sufficient to account for most or all teocytes was much higher than of osteoblasts, reflecting the
of those missing (8). Based on the dynamic histomorphometry unique unavailability of osteocytes for phagocytosis because of
at the murine vertebral secondary spongiosa and a wall width their anatomic isolation from scavenger cells, and the need for
of z 15 mm (7, 9, 14), we calculated the mean active life span extensive degradation to small molecules to dispose of the cells
of an osteoblast on cancellous bone by dividing wall width by through the narrow canaliculi. As a result, the process is pro-
the mineral appositional rate. From this calculation, we esti- longed and affected cells accumulate.
mated that the mean active life span of a murine osteoblast is The contribution of osteocyte apoptosis to glucocorticoid-
z 12 d or 288 h. The prevalence of osteoblast apoptosis in this induced bone disease will remain uncertain until more is
study was 0.0066 in the placebo group. We applied the follow- known about the function of these cells, but two possibilities
ing relationship: merit discussion. First, the network of osteocytes probably par-
ticipates in the detection of microdamage and the transmission
t Ap /288 = 0.0066/ f Ap ;
of signals that lead to its repair by remodeling (50). Disruption
where tAp is the mean duration (in hours) of the DNA frag- of the network by osteocyte apoptosis could compromise this
mentation phase of apoptosis that is detected by TUNEL, and mechanism, leading to microdamage accumulation and in-
fAp is the fraction of osteoblasts that undergoes apoptosis. creased bone fragility (51). Second, chronic glucocorticoid ad-
Based on a value of tAp of z 3 h, determined previously for re- ministration sometimes leads to so-called aseptic or avascular
generating liver (44), the corresponding value for fAp in the pla- necrosis of bone, a painful and disabling complication (6). The
cebo group is 0.6. The calculation demonstrates that the low name may be misleading, since it has not been demonstrated
prevalence of apoptosis we observed in the placebo group is that the bone cells die by necrosis. Indeed, the cell swelling and
consistent with the conclusion drawn from studies of human tissue inflammation that characterize necrosis in soft tissues do
bone that 50–70% of osteoblasts undergo apoptosis, and that not occur (6, 52). Glucocorticoid-induced osteocyte apoptosis,
only a minority become osteocytes or lining cells (43). a cumulative and unrepairable defect, would explain the corre-
In the animals receiving the higher dose of prednisolone, lation between total dose and incidence of avascular necrosis
the prevalence of apoptosis was 0.0203. With prednisolone ad- of bone (53) and its occurrence after glucocorticoid adminis-
ministration, phagocytosis of the apoptotic cells would be sup- tration had ceased.
pressed and we estimated that tAp could be doubled (45). Wall In conclusion, we have demonstrated that the mouse is a
width was reduced to z 8 mm and mineral appositional rate to valid and informative model of glucocorticoid-induced bone
0.74 mm/d, so that the active life span of an osteoblast is z 260 h. disease, not confounded by weight loss or sex steroid defi-
In these circumstances, the corresponding value for fAp in the ciency, and that many of the effects of chronic glucocorticoid
prednisolone group is 0.9. Although there is some uncertainty administration on bone can be explained by decreased birth of
to the assumptions used for these estimates, the approach does osteoblast and osteoclast precursors and increased apoptosis
help explain the data and disclose the devastating impact of of mature osteoblasts and osteocytes.
glucocorticoid excess on osteoblast survival. The higher pro-
portion of osteoblasts showing features of apoptosis in gluco- Note added in proof: While our manuscript was in press, Rei-
corticoid-treated mice and human subjects could indicate no chardt et al. reported that glucocorticoid-induced apoptosis is medi-
more than prolongation of the time needed for completion of ated by a mechanism that requires the dimerization of the glucocorti-
the process, but we think it more likely that glucocorticoids in- coid receptor and direct binding of the receptor to GRE-response
duce apoptosis, either prematurely in cells already destined for elements (1998. Cell. 93:531–541). Taken together with the develop-
this fate or in cells otherwise destined to become lining cells or ment of synthetic glucocorticoids that exhibit antiinflammatory activ-
ity in vivo as potently as classical glucocorticoids, without requiring
osteocytes. In either case, the mean active life span of osteo-
GR-DNA-binding and transactivation (1997. Mol. Endocrinol. 11:
blasts would be shortened and less bone formed. Thus, the 1245–1255), the present demonstration of osteoblast and osteocyte
demonstrated reduction in bone formation by glucocorticoids apoptosis in animals and humans with glucocorticoid-induced os-
could be due to increased death as well as decreased birth of teoporosis predicts that these synthetic compounds will have bone
osteoblasts. Further support of this concept is given by the re- sparing effects.
cent report of glucocorticoid-induced apoptosis of osteoblasts
in calvariae of young mice given dexamethasone (46).
Acknowledgments
Osteocytes are long-lived but not immortal cells. In human
rib cortical bone, their life span has been estimated at z 20 yr The authors wish to acknowledge Julie Crawford, Frances Swain,
(47); if bone remains unremodeled for a longer time, the os- Carman Young, James Kirchner, Randal Shelton, Sherry Rush, and
teocytes die, as revealed by empty lacunae and hypermineral- Catherine Smith for their invaluable technical assistance in the con-
ized perilacunar bone, referred to as micropetrosis (48). Os- duct of these studies.

Regulation of the Birth and Death of Osteoblasts by Glucocorticoids 281

 This work was supported by the National Institutes of Health 26. Grardel, B., B. Sutter, B. Flautre, E. Viguier, F. Lavaste, and P. Har-
(PO1AG13918 and RO1AR43003) and the Department of Veterans douin. 1994. Effects of glucocorticoids on skeletal growth in rabbits evaluated
by dual-photon absorptiometry, microscopic connectivity and vertebral com-
Affairs.
pressive strength. Osteoporosis Int. 4:204–210.
27. Kawai, K., A. Tamaki, and K. Hirohata. 1985. Steroid-induced accumu-
lation of lipid in the osteocytes of the rabbit femoral head. J. Bone Joint Surg.
References 67A:755–762.
28. Deloffre, P., D. Hans, C. Rumelhart, D. Mitton, Y. Tsouderos, and P.J.
1. Cushing, H. 1932. The basophil adenomas of the pituitary body and their Meunier. 1995. Comparison between bone-density and bone strength in gluco-
clinical manifestations (pituitary basophilism). Bull. Johns Hopkins Hosp. 50: corticoid-treated aged ewes. Bone (Tarrytown). 17:409S–414S.
137–195. 29. Chavassieux, P., A. Buffet, P. Vergnaud, P. Garnero, and P.J. Meunier.
2. Lukert, B. 1996. Glucocorticoid-induced osteoporosis. In Osteoporosis. 1997. Short-term effects of corticosteroids on trabecular bone remodeling in old
R. Marcus, D. Feldman, and J. Kelsey, editors. Academic Press, San Diego, ewes. Bone (Tarrytown). 20:451–455.
CA. 801–820. 30. Newman, E., A.S. Turner, and J.D. Wark. 1995. The potential of sheep
3. Fitzpatrick, L.A. 1994. Glucocorticoid-induced osteoporosis. In Os- for the study of osteopenia: current status and comparison with other animal
teoporosis. R. Marcus, editor. Blackwell Scientific Publications, Boston, MA. models. Bone (Tarrytown). 16:277S–284S.
202–226. 31. Dempster, D.W., B.S. Moonga, L.S. Stein, W.R. Horbert, and T.
4. Reid, I.R. 1989. Pathogenesis and treatment of steroid osteoporosis. Clin. Antakly. 1997. Glucocorticoids inhibit bone resorption by isolated osteoclasts
Endocrinol. 30:83–103. by enhancing apoptosis. J. Endocrinol. 154:397–406.
5. Dempster, D. 1989. Bone histomorphometry in glucocorticoid-induced 32. Conaway, H.H., D. Grigorie, and U.H. Lerner. 1996. Stimulation of
osteoporosis. J. Bone Miner. Res. 4:137–141. neonatal mouse calvarial bone resorption by the glucocorticoids hydrocortisone
6. Mankin, H.J. 1992. Nontraumatic necrosis of bone (osteonecrosis). N. and dexamethasone. J. Bone Miner. Metab. 11:1419–1429.
Engl. J. Med. 326:1473–1479. 33. Bellows, C.G., J.E. Aubin, and J.N.M. Heersche. 1987. Physiological
7. Jilka, R.L., R.S. Weinstein, K. Takahashi, A.M. Parfitt, and S.C. Manola- concentrations of glucocorticoids stimulate formation of bone nodules from iso-
gas. 1996. Linkage of decreased bone mass with impaired osteoblastogenesis in lated rat calvarial cells in vitro. Endocrinology. 121:1985–1992.
a murine model of accelerated senescence. J. Clin. Invest. 97:1732–1740. 34. Lian, J.B., V. Shalhoub, F. Aslam, B. Frenkel, J. Green, M. Hamrah,
8. Jilka, R.L., R.S. Weinstein, T. Bellido, A.M. Parfitt, and S.C. Manolagas. G.S. Stein, and J.L. Stein. 1997. Species-specific glucocorticoid and 1,25-dihy-
1998. Osteoblast programmed cell death (apoptosis): modulation by growth droxyvitamin D responsiveness in mouse MC3T3-E1 osteoblasts: dexametha-
factors and cytokines. J. Bone Miner. Res. 13:793–802. sone inhibits osteoblast differentiation and vitamin D down-regulates osteocal-
9. Weinstein, R.S., R.L. Jilka, A.M. Parfitt, and S.C. Manolagas. 1997. The cin gene expression. Endocrinology. 138:2117–2127.
effects of androgen deficiency on murine bone remodeling and bone mineral 35. Stanton, B., G. Giebisch, G. Klein-Robbenhaar, J. Wade, and R.A. De-
density are mediated via cells of the osteoblastic lineage. Endocrinology. 138: Fronzo. 1985. Effects of adrenalectomy and chronic corticosteroid replacement
4013–4021. on potassium transport in rat kidney. J. Clin. Invest. 75:1317–1326.
10. Frey, F.J. 1987. Kinetics and dynamics of prednisolone. Endocr. Rev. 8: 36. Borchard, R.E., C.D. Barnes, and L.G. Eltherington. 1992. Drug Dos-
453–473. age in Laboratory Animals: A Handbook. CRC Press, Inc., Boca Raton, FL.
11. Cope, C.L. 1972. The synthetic analogues. In Adrenal Steroids and Dis- 514–517.
ease. Lippincott, Philadelphia, PA. 488–491. 37. Kleiber, M. 1961. The Fire of Life: An Introduction to Animal Energet-
12. Broulik, P.D., and L. Stárka. 1997. Effect of antiandrogens casodex and ics. John Wiley & Sons, Inc., New York. 177–230.
epitestosterone on bone composition in mice. Bone (Tarrytown). 20:473–475. 38. Klein, M., A.R. Villanueva, and H.M. Frost. 1965. A quantitative histo-
13. Weinstein, R.S., and N.H. Bell. 1988. Diminished rates of bone forma- logical study of rib from 18 patients treated with adrenal cortical steroids. Acta
tion in normal black adults. N. Engl. J. Med. 319:1698–1701. Orthop. Scand. 35:171–184.
14. Parfitt, A.M., M.K. Drezner, F.H. Glorieux, J.A. Kanis, H. Malluche, 39. Bressot, C., P.J. Meunier, M.C. Chapuy, E. Lejeune, C. Edouard, and
P.J. Meunier, S.M. Ott, and R.R. Recker. 1987. Bone histomorphometry: stan- A.J. Darby. 1979. Histomorphometric profile, pathophysiology and reversibility
dardization of nomenclature, symbols, and units. Report of the ASBMR Histo- of corticosteroid-induced osteoporosis. Metab. Bone Dis. Relat. Res. 1:303–311.
morphometry Nomenclature Committee. J. Bone Miner. Res. 2:595–610. 40. Parfitt, A.M., A.R. Villanueva, J. Foldes, and D.S. Rao. 1995. Relations
15. Parfitt, A.M., C.H.E. Mathews, A.R. Villanueva, M. Kleerekoper, B. between histologic indices of bone formation: implications for the pathogenesis
Frame, and D.S. Rao. 1983. Relationships between surface, volume and thick- of spinal osteoporosis. J. Bone Miner. Res. 10:466–473.
ness of iliac trabecular bone in aging and in osteoporosis. Implications for the 41. Parfitt, A.M. 1994. Osteonal and hemi-osteonal remodeling: the spatial
microanatomic and cellular mechanism of bone loss. J. Clin. Invest. 72:1396– and temporal framework for signal traffic in adult human bone. J. Cell. Bio-
1409. chem. 55:273–286.
16. Bellows, C.G., J.E. Aubin, and J.N.M. Heersche. 1991. Initiation and 42. Manolagas, S.C., R.L. Jilka, T. Bellido, C.A. O’Brien, and A.M. Parfitt.
progression of mineralization of bone nodules formed in vivo: the role of alka- 1996. Interleukin-6-type cytokines and their receptors. In Principles of Bone Bi-
line phosphatase and organic phosphate. Bone Miner. 14:27–40. ology. J.P. Bilezikian, L.G. Raisz, and G.A. Rodan, editors. Academic Press,
17. Owen, M. 1985. Lineage of osteogenic cells and their relationship to the San Diego, CA. 701–713.
stromal system. In Bone and Mineral Research. Volume 3. W.A. Peck, editor. 43. Parfitt, A.M. 1990. Bone-forming cells in clinical conditions. In Bone: A
Elsevier, Amsterdam. 1–25. Treatise, Vol. 1. The Osteoblast and Osteocyte. B.K. Hall, editor. Telford and
18. Falla, N., P. Van Vlasselaer, J. Bierkens, B. Borremans, G. Schoeters, CRC Press, Boca Raton, FL. 351–429.
and U. Van Gorp. 1993. Characterization of a 5-fluorouracil-enriched osteopro- 44. Bursch, W., S. Paffe, B. Putz, and R. Schulte-Hermann. 1990. Determi-
genitor population of the murine bone marrow. Blood. 82:3580–3591. nation of the length of the histological stages of apoptosis in normal liver and in
19. Tomkinson, A., J. Reeve, R.W. Shaw, and B.S. Noble. 1997. The death altered hepatic foci of rats. Carcinogenesis. 11:847–583.
of osteocytes via apoptosis accompanies estrogen withdrawal in human bone. J. 45. Cline, M.J. 1974. Drugs and phagocytes. N. Engl. J. Med. 291:1187–1188.
Clin. Endocrinol. Metab. 82:3128–3135. 46. Gohel, A., and G. Gronowicz. 1997. Glucocorticoids induce apoptosis in
20. Lynch, M.P., C. Capparelli, J.L. Stein, G.S. Stein, and J.B. Lian. 1998. osteoblasts in mice by the regulation of BCL-2, BAX and other cell cycle fac-
Apoptosis during bone-like tissue development in vitro. J. Cell. Biochem. 68: tors. J. Bone Miner. Res. 12(Suppl. 1):S284. (Abstr.)
31–49. 47. Frost, H.M. 1960. In vivo osteocyte cell death. J. Bone Joint Surg. 42A:
21. StatCorp. 1995. Stata Statistical Software Release 4.0. Stata Corpora- 138–143.
tion, College Station, TX. 1–1601. 48. Frost, H.M. 1960. Micropetrosis. J. Bone Joint Surg. 42A:144–150.
22. Jilka, R.L., G. Hangoc, G. Girasole, G. Passeri, D.C. Williams, J.S. 49. Dunstan, C.R., N.M. Somers, and R.A. Evans. 1993. Osteocyte death
Abrams, B. Boyce, H. Broxmeyer, and S.C. Manolagas. 1992. Increased osteo- and hip fracture. Calcif. Tissue Int. 53:S113–S117.
clast development after estrogen loss: mediation by interleukin-6. Science. 257: 50. Aarden, E.M., E.H. Burger, and P.J. Nijweide. 1994. Function of osteo-
88–91. cytes in bone. J. Cell. Biochem. 55:287–299.
23. King, C.S., E.C. Weir, C.W. Gundberg, J. Fox, and K.L. Insogna. 1996. 51. Noble, B.S., H. Stevens, N. Loveridge, and J. Reeve. 1997. Identification
Effects of continuous glucocorticoid infusion on bone metabolism in the rat. of apoptotic osteocytes in normal and pathological human bone. Bone (Tarry-
Calcif. Tissue Int. 59:184–191. town). 20:273–282.
24. Li, M., Y. Shen, B.P. Halloran, B.D. Baumann, K. Miller, and T.J. 52. Ficat, R.P. 1985. Idiopathic necrosis of the femoral head. J. Bone Joint
Wronski. 1996. Skeletal response to corticosteroid deficiency and excess in Surg. 67B:3–9.
growing male rats. Bone (Tarrytown). 19:81–88. 53. Felson, D.T., and J.J. Anderson. 1987. Across-study evaluation of asso-
25. Quarles, L.D. 1992. Prednisone-induced osteopenia in beagles: variable ciation between steroid dose and bolus steroids and avascular necrosis of bone.
effects mediated by differential suppression of bone formation. Am. J. Physiol. Lancet. I:902–905.
(Endocrinol. Metab.). 263:E136–E141.

282 Weinstein et al.