Full text of DOI:10.1359/jbmr.2001.16.2.328

JOURNAL OF BONE AND MINERAL RESEARCH
Volume 16, Number 2, 2001
©
2001 American Society for Bone and Mineral Research
T Cell Activation Induces Human Osteoclast Formation via
Receptor Activator of Nuclear Factor ␬B Ligand–
Dependent and –Independent Mechanisms
1
,2
1,2
1
3
M. NEALE WEITZMANN, SIMONE CENCI, LEONARD RIFAS, JEFFREY HAUG,
3
1
JOHN DIPERSIO, and ROBERTO PACIFICI
ABSTRACT
In unstimulated conditions, osteoclast (OC) formation is regulated by stromal cell production of the key
osteoclastogenic factors receptor activator of nuclear factor ␬B ligand (RANKL) and macrophage colony–
stimulating factor (M-CSF). However, the mechanisms of accelerated osteoclastogenesis and bone loss
characteristic of inflammatory conditions are poorly understood but appear to involve T cells. In addition, the
mechanism by which OCs arise spontaneously in cultures of peripheral blood mononuclear cells in the absence
of stromal cells or added cytokines remains unclear. Using a stromal cell free human osteoclast generating
system, we investigated the ability of activated T cells to support osteoclastogenesis. We show that when
؉
؉
activated by phytohemagglutinin-P (PHA), T cells (both CD4 and CD8 ) stimulate human OC formation in
vitro. Although both soluble M-CSF and RANKL were detected in activated T cell supernatants, the presence
of M-CSF was not essential for macrophage survival or RANKL-dependent osteoclast formation, suggesting
that other soluble T cell–derived factors were capable of substituting for this cytokine. We also found that
saturating concentrations of osteoprotegerin (OPG) failed to neutralize 30% of the observed OC formation
and that T cell conditioned medium (CM) could superinduce osteoclastogenesis in cultures of purified
monocytes maximally stimulated by RANKL and M-CSF. Together, these data suggest that activated T cells
support osteoclastogenesis via RANKL-dependent and -independent mechanisms. Although not relevant for T
cell–induced osteoclastogenesis, secretion of soluble M-CSF is a previously undescribed property of activated
T cells. (J Bone Miner Res 2001;16:328–337)
Key words: T cell, osteoclast, receptor activator of nuclear factor ␬B ligand, macrophage colony–stimulating
factor, osteoprotegerin
(
1)
INTRODUCTION
of T cells in peripheral blood, an increase in suppressor T
(2)
ϩ
ϩ
(3)
cell activity, and elevation of CD4 /CD8 ratios have
T HAS LONG been recognized that osteoporosis in humans all been reported in osteoporotic patients. Activated T cells
is associated with alterations in human T lymphocyte also have been associated with increased osteoclast (OC)
populations. An increase in the relative and absolute number formation and accelerated bone resorption under inflamma-
I
1
Division of Bone and Mineral Diseases, Washington University School of Medicine and Barnes-Jewish Hospital, St. Louis, Missouri,
USA.
2
These authors contributed equally to this work.
Division of Bone Marrow Transplant and Stem Cell Biology, Washington University School of Medicine and Barnes-Jewish Hospital,
3
St. Louis, Missouri, USA.
3
28

T-CELLS INDUCE OSTEOCLASTOGENESIS
329
(4,5)
(6)
tory conditions in vivo
and in vitro. The T cell mito- 121F was a kind gift from Dr. P. Osdoby (Washington
gen phytohemagglutinin-P (PHA) has been reported to stim- University). OPG, anti–M-CSF, and anti-RANKL antibod-
(7,8)
ulate OC formation and bone resorption
presumably ies were from R&D Systems (Minneapolis, MN, USA).
through T cell–mediated events. Whether the observed ef-
fects of PHA on OC formation are mediated by T lympho-
cytes, monocytes, or other cell types present in culture
Generation of human OCs
(
16)
Human OCs were generated as previously described.
ϩ
remains unknown. The ability of activated T cells to support Briefly, human PBSCs, a blood product enriched in CD34
(9,10)
osteoclastogenesis
is thought to be caused by expres- stem cells, were prepared by a single leukapheresis of
sion of the key osteoclastogenic cytokine receptor activator healthy donors, treated with G-CSF for 3–5 days. After
(11)
of nuclear factor ␬B ligand (RANKL), also known as OC purification of mononuclear cells by Histopaqe density gra-
(
12)
differentiation factor (ODF),
osteoprotegerin ligand dient centrifugation, PBSCs were plated at a density of 5 ϫ
(13)
5
(OPGL),
and tumor necrosis factor–related activation– 10 cells/well in a final volume of 0.5 ml using 48-well
(14)
induced cytokine (TRANCE).
plates containing 10-mm wells. Cells were cultured in ␣
It also has been established that OCs can arise spontane- modified essential medium (␣-MEM) supplemented with
ously at low frequency in normal human peripheral blood 10% fetal bovine serum (FBS; Gibco BRL, Gaithersburg,
mononuclear cell cultures in the absence of stromal cells or MD, USA), penicillin (50 U/ml), and streptomycin (50
exogenous cytokines. However, the mechanism driving for- ␮g/ml) for 21 days in the presence of IL-1, IL-3, and
mation of these OCs remains unknown but has been sug- GM-CSF (10 ng/ml each) or for 7 days in the presence of
gested to be via RANKL production by adherent cells in the PHA (1.5 ␮g/ml), in the absence of other exogenous cyto-
(
15)
culture.
kines. Cells were fed by replacing half the medium twice
ϩ
Previously, we have indicated that human CD34 mobi- weekly with addition of fresh cytokines. PHA was only
lized peripheral blood stem cells (PBSCs) form large added once at time of seeding.
numbers of authentic OCs after 21 days in culture in the
For some experiments, purified populations of monocytes
5
ϩ ϩ
presence of interleukin (IL)-1, IL-3, and granulocyte- (1 ϫ 10 cells/well) and CD4 and CD8 subpopulations of
(16)
5
macrophage colony–stimulating factor (GM-CSF).
We T cells (1 ϫ 10 cell/well) were combined in a final volume
now show that PHA activation of T cells present in PBSC of 0.5 ml using 48-well plates. Some wells of purified
5
cultures results in the rapid formation (7 days) of large monocytes (1 ϫ 10 cells/well) received T cell CM, which
5
numbers of mature OCs, in the absence of exogenous cyϩ- was derived from a 4ϫ concentration of T cells (4 ϫ 10
tokines. Our data show that activated T cells, both CD4
cells/500 ␮l) and added at 25% final volume (125 ␮l)
ϩ
and CD8 , secrete soluble osteoclastogenic factors includ- representing a 1ϫ final dose.
ing a secreted form of RANKL, which is responsible for
ϳ70% of the osteoclastogenic activity of T cell conditioned
medium (CM) in this system. However, an additional ac-
tivity, accounting for 30% of OC formation was found to be
Isolation/depletion of specific cell populations
T cell and monocyte (early OC precursor) populations
RANKL independent. Although soluble M-CSF was de- were isolated or depleted from PBSC cultures using anti-
tected in T cell CM, M-CSF neutralization experiments CD14 (monocytes/macrophages), anti-CD3 (T cells), and
showed that M-CSF is not required in this system for anti-CD4 and CD8 (T cell subset), coated immunomagnetic
RANKL activity.
Dynabeads (Dynal Lake Success, NY, USA), according to
These findings suggest a dual mechanism for T cell– the manufacturer’s instructions. T cells positively selected
mediated osteoclastogenesis involving RANKL-dependent using Dynabeads are 100% pure and Ͼ95% viable afteϩr
and -independent mechanisms and the presence of factors isolation. Dynabeads were detached from purified CD3
ϩ
ϩ
capable of substituting for M-CSF.
T cells, after 6 h, by gentle pipetting. For CD4 and CD8
T cell subsets, attached beads were removed using DE-
TACHaBEAD (Dynal) according to the manufacturer’s in-
structions. The CD4 and CD8 antibodies used in the T
cell isolation procedure do not activate the cells; however,
ϩ
ϩ
MATERIALS AND METHODS
All reagents were purchased from the Sigma Chemical the CD3 antibody is known to activate the T cells. Thus for
St. Louis, MO, USA), unless otherwise indicated. The some experiments, purified unactivated T cells were ob-
(
recombinant human IL-1␤ (rhIL-1␤) was kindly provided tained using negative selection columns (R&D Systems). T
by Dompe Pharmaceutical (L’Aquila, Italy) and rhIL-3, cell purity of Ն95% (as accessed by flow cytometry), was
rhM-CSF, and rhGM-CSF were provided by Genetics In- achieved by predepletion of adherent cell populations
stitute (Boston, MA, USA). Polyclonal antibody directed (monocytes and B cells) by 3-h adhesion to plastic, followed
against the human integrin subunits ␣V (clone L230) and by double negative selection on T cell selection columns.
␤
3 (clone AP3) were generous gifts of Dr. S.D. Blystone
As immunomagnetic beads are phagocytosed by mono-
(Washington University, St. Louis, MO, USA). K10C7, an cytes and cannot be removed, monocytes (a source of early
affinity purified murine monoclonal antibody recognizing OC precursors) were semipurified, after immunomagnetic
the mature form of Cathepsin K was a kind gift of Dr. M. antibody depletion of B cells (an adherent cell population),
Gowen, SmithKline Beecham, (King of Prussia, PA, USA). by adhesion onto plastic dishes for 3 h, and nonadherent
c-Src
Anti-pp60
(clone 327) was kindly provided by Dr. Y. cells were removed by washing three times with phosphate-
Abu-Amer (Washington University). The anti-OC antibody buffered saline (PBS). These populations were found to

3
30
WEITZMANN ET AL.
respond in the same manner as monocytes positively se- containing FBS to stop the enzymatic reaction. Cells were
lected by CD14 immunomagnetic beads and were found to recovered by centrifugation and seeded onto whale dentine
ϩ
be Ն64% CD14 and Ն99% T cell and B cell depleted as slices in 48-well plates. After 2–5 days of culture, slices
assessed by flow cytometry.
were washed with PBS and incubated for 18 h in 0.25 M
ammonium hydroxide followed by sonication in PBS for
3
0 s. Dentine slices were washed with PBS and stained in
Characterization of human OCs
toluidine blue (2% wt/vol) in PBS for 5 minutes. Pits were
Tartrate-resistant acid phosphatase: After 7 days of cul- photographed on an Olympus phase contrast inverted mi-
ture in the presence of PHA, cultures were stained for croscope (Olympus, Tokyo, Japan).
tartrate-resistant acid phosphatase (TRAP) by trypsinizing
wells for 5–10 minutes to remove early (monocytes) and
late OC precursors and other loosely adhered cells in cul-
ture. The remaining cells were stained for TRAP using a
commercial kit (Sigma) according to the manufacturer’s
instructions. OCs were found to adhere extremely tightly to
the plastic and are not removed by treatment with trypsin.
M-CSF ELISA
M-CSF was measured after a 48-h incubation in CM from
unstimulated and PHA (1.5 ␮g/ml)-stimulated T cells, pu-
rified from PBSCS, using a commercial ELISA for rhM-
CSF (Quantikine; R&D Systems). T cell supernatants from
three independent experiments were measured in duplicate
and data were expressed as the mean concentration of
Immunohistochemistry: Immunohistochemical staining
c-src
for ␣V- and ␤3-integrin subunits pp60
cathepsin K and
6
M-CSF (ng/ml/10 cells) Ϯ SD.
for antigen to the anti-OC antibody 121F was conducted as
(17)
previously described.
Briefly, cells were grown in 48-
well plates and washed three times with PBS before fixa- M-CSF neutralization
tion. Cells were fixed in acetone/methanol (50%:50%, vol/
For M-CSF neutralization experiments, polyclonal M-CSF
vol) for 5 minutes and washed twice with PBS. Fixed cells
were incubated with primary antibody (5 ␮g/ml) in PBS
containing 3% bovine serum albumin (BSA) at 4° over-
night. Wells were washed three times in PBS for 5 minutes
with shaking. Cells were incubated with secondary antibody
antibody (20 ␮g/ml) was added to wells before addition of T
cell CM. This concentration of antibody was found to neutral-
ize completely OC formation induced by RANKL and 500
ng/ml of rhM-CSF, a dose 100-fold higher than the concen-
tration of soluble M-CSF, was estimated to be present in T cell
CM by ELISA. In addition, monocytes treated with M-CSF
and anti–M-CSF antibody were completely dead after 7 days
in culture compared with Ͼ95% viability with M-CSF–
stimulated cells alone, assessed by trypan blue dye exclusion.
(anti-mouse immunoglobulin G [IgG]-biotin conjugate), in
PBS containing 3% BSA for 2 h at room temperature,
washed three times, and incubated with avidin-peroxidase,
in PBS containing 3% BSA for 40 minutes at room tem-
perature. After three washes, color was developed using
4
1
-chloronapthol (0.03% in 0.05 M Tris-HCl, pH 7.6, and 0.
% H O ). Wells were washed with water after several
Semiquantitative reverse-transcription polymerase
chain reaction for RANKL
2
2
minutes to stop the staining. Strong staining in relation to
the IgG isotype control was considered positive. No sub-
stantial staining was observed in the absence of primary
antibody showing that endogenous peroxidase did not con-
tribute to the signal.
Total RNA from unstimulated and PHA-stimulated T
cells was isolated using Trizol Reagent (Gibco BRL) ac-
cording to the manufacturer’s instructions. Semiquantitative
reverse-transcription polymerase chain reaction (RT-PCR)
was conducted using Ready-To-Go RT-PCR beads (Amer-
sham Pharmacia Biotech, Piscataway, NJ, USA) according
to the manufacturer’s instructions. Briefly, 2 ␮g of total
RNA were reverse transcribed using a pd(T)12–18 first-strand
primer for 30 minutes at 42°C, before PCR. The for-
ward RANKL primer used was 5Ј-ATGCGCCGCGCC-
AGCAGAGACTACA, and the reverse primer 5Ј-ATC-
TATATCTCGAACTTTAAAAGCC. Commercial human
glyceraldehyde-3-phosphate dehydrogenase (GAPDH) am-
plimer primers (Clontech, Palo Alto, CA, USA) were used
as internal standards. Genomic contamination was assessed
by heat inactivation of the reverse transcriptase before RT-
PCR. RANKL was subjected to 40 cycles of PCR and found
to be in the exponential amplification range up to 45 cycles.
GAPDH was cycled for 30 rounds and was exponential up
to 35 cycles.
Calcitonin receptor analysis: Expression of calcitonin
1
25
receptors was assessed by autoradiography using
labeled salmon calcitonin (2000 Ci/mmol; Amersham Inter-
national, Arlington Heights, IL, USA), as previously de-
scribed.
previously, washed twice in PBS to remove nonadherent
cells, and labeled with 0.2 nM I-calcitonin for 1 h at room
temperature. Two hundred–fold excess unlabeled salmon
calcitonin was used to assess nonspecific binding. After
washing, the cells were fixed, stained for TRAP, and coated
with LM-1 photographic emulsion (Amersham Interna-
tional) and stored for 3 weeks at 4°C before developing.
Bone resorption pit assays: The ability of OCs to resorb
pits in bone was assessed by generating mature OCs and
transplanting mature OCs onto dentine slices as previously
I-
(16)
Briefly, mature OCs were cultured as described
1
25
(16)
described.
Briefly, cell layers were rinsed with
Ca ,Mg -free Tyrode’s solution, incubated at 37°C for
5 minutes with the same buffer, and then incubated in
2
ϩ
2ϩ
1
Metabolic labeling for detection of soluble RANKL
2
ϩ
2ϩ
Ca ,Mg -free Tyrode’s solution with 1 mM EDTA and
4 mg/ml trypsin at 37°C for 20 minutes. Cells were then
4
T cells were labeled metabolically as previously de-
(18)
removed by agitation with a pipette and transferred to a tube scribed.
Briefly, cells were labeled overnight with 50

T-CELLS INDUCE OSTEOCLASTOGENESIS
Ci of EXPRESS label (Amersham Pharmacia Biotech)
331
␮
and then stimulated for 24 h with PHA. Supernatant was
immunoprecipitated with antibody against human RANKL
and protein G beads. The immunoprecipitated material was
recovered by boiling in sample loading buffer and separated
by sodium dodecyl sulfate–polyacrylamide gel electro-
phoresis (SDS-PAGE). Gels were dried and exposed to
autoradiography film (Kodak XAR; Eastman Kodak Co.,
Rochester, NY, USA).
Statistical analysis
Group mean values were compared by unpaired two-
tailed Student’s t-test or one-way analysis of variance
(ANOVA) as appropriate. Subsequent multiple compari-
sons were performed by the Fisher Protected least signifi-
cant difference test.
FIG. 1. PHA stimulation accelerates PBSC OC formation. PBSC
cultures stimulated with PHA in the absence of exogenous cytokines
generate OCs within 7 days compared with cultures stimulated with
IL-1, IL-3, and GM-CSF for 21 days in the absence of PHA. Each data
point represents the average Ϯ SD of triplicate wells. Similar results
RESULTS
Human PBSCs are a mixed population of mononuclear
cells containing monocytes/macrophages, early OC precur-
sors of the monocytic lineage, and T and B lymphocytes. were obtained in three independent experiments.
PBSCs do not contain stromal cells or stromal cell precur-
sors. When cultured in vitro with IL-1, IL-3, and GM-CSF,
PBSCs generate large numbers of authentic bone resorbing
To confirm that the stimulatory effects of PHA on OC
(16)
OCs within 21 days of culture.
In the current study we maturation were mediated through T lymphocytes, we im-
ϩ
have used this stromal cell free human model system to munomagnetically depleted mature CD3 T cells from OC
evaluate the contribution of activated T cells to the process cultures. The data shows (Fig. 3) that in the absence of T
of osteoclastogenesis. To achieve this aim, PBSCs were cells, the ability of PHA to stimulate OCs within a 7-day
cultured with or without the T cell activator PHA. In the period is reduced by 3-fold. Importantly, addition of nega-
absence of exogenous cytokines, activated T cells generated tively isolated (unstimulated) T cells back into T cell–
mature OCs within only 7 days of culture compared with 21 depleted cultures restored the osteoclastogenic response to
days required in the absence of stimulated T cells (Fig. 1). PHA.
OCs formed at 7 days under PHA-stimulated conditions
died over the subsequent 2 weeks resulting in lower num- precursors or require other intermediary cells, T cells were
bers by 21 days. purified from PBSCS by negative selection columns and
To determine if activated T cells act directly on OC
The total number of OCs formed in the presence of PHA added to monocytes purified from PBSCS (early OC pre-
at 7 days was not appreciably different from the number cursors). Cultures were incubated in the presence or absence
forming in the absence of PHA over a 21-day period (Fig. of PHA. PHA-stimulated T cells were found to stimulate
1
). This suggests that PHA affects OC generation rather directly OC production in the absence of other supporting
than proliferation of precursors. cells (Fig. 4.). Purified T cells or monocytes alone failed to
TRAP cells, formed in 7 days after PHA stimulation, induce OC formation.
ϩ
showing Ն3 nuclei were defined as mature OCs and were
To evaluate which populations of T cells support osteo-
characterized for multiple specific markers of OCs. Consis- clastogenesis, we used a positive immunomagnetic selec-
(
16)
ϩ
ϩ
tent with previous observations,
we have found that tion to purify either CD4 or CD8 T cells from PBSC
ϩ
Ͼ90% of TRAP multinucleated cells (Ն3 nuclei) gener- cultures. This selection procedure does not activate the T
ated in these studies exhibit calcitonin receptor, a specific cells. T cell subtypes were then added to purified PBSC
marker of human OCs. In addition, we have found that monocytes and the cultures were treated with or without
calcitonin receptor expression correlates (r ϭ 0.98) ex- PHA. The data (Fig. 5A) show that both activated CD4 or
tremely well with cathepsin K expression, another specific CD8 T cells were equally capable of inducing OCs from
ϩ
ϩ
marker of OCs. Figure 2 shows the ability of multinucleated cultures of PBSC monocytes. In addition, CM alone from
ϩ
ϩ
cells generated with PHA in this culture system to resorb cultures of activated CD4 or CD8 T cell subsets (Fig. 5B)
pits in dentine slices, express calcitonin receptors, integrin was capable of stimulating OC formation from PBSC
c-src
subunits ␣V and ␤3, pp60
mature cathepsin K, and to monocytes, suggesting the presence of soluble osteoclasto-
react with the antibody 121F, which recognizes avian and genic factors. The addition of PHA alone to monocytes or
(19)
ϩ
ϩ
human OCs. Based on these criterion, these cells meet all medium from unstimulated CD4 or CD8 T cell subsets
current requirements for authentic human OCs. failed to induce substantial OC formation (Figs. 5A and

3
32
WEITZMANN ET AL.
FIG. 2. Characterization of human OCs gen-
erated by PHA stimulation of PBSCs. The
multinuclear cells formed in 7 days in the pres-
ence of PHA were analyzed for markers of au-
thentic human OCs (see Materials and Methods
section). Multinuclear cells were found to (A)
resorb pits on dentine slices; (B) express calci-
tonin receptors that were competed out (C) by
2
00-fold excess unlabeled calcitonin; (B and C)
TRAP (pink coloration on multinuclear cells,
note unstained mononuclear precursors); (D) ex-
press the integrin subunits ␣V (detected with
antibody L230) and (E) ␤3 detected with anti-
c-src
body AP3; express (F) pp60
(detected with
antibody 327), (G) mature cathepsin K (detected
with antibody K10C7) and (H) react with the
anti-OC antibody 121F. (I) Nonspecific binding
was assessed by IgG isotype control. Panels A
and D–I were photographed at magnification
ϫ200, and panels B and C were photographed at
magnification ϫ300.
5
B). The same results were observed using T cells isolateϩd
To assess directly whether the enhanced osteoclastogenic
activity induced by activated T cell CM was caused by the
presence of secreted RANKL and M-CSF, we added the
from normal human peripheral blood as opposed to CD34
enriched PBSCs.
To investigate possible mechanisms by which PHA- RANKL inhibitor OPG or neutralizing anti–M-CSF anti-
activated T cells could induce OC formation, we evaluated body into cultures of monocytes stimulated with PHA CM
the ability of PHA to induce the key osteoclastogenic fac- (1:4 final volume). Saturating concentrations of OPG (1
tors M-CSF and RANKL from T cells isolated from PBSCs. ␮g/ml) were found to decrease OC formation in PHA-
To investigate this possibility, we conducted ELISAs to stimulated cultures by ϳ3-fold (Fig. 7). Because higher
measure M-CSF in supernatants of T cell CM. PHA (1.5 concentrations of OPG, up to 4 ␮g/ml, failed to further
␮
5
g/ml) stimulation of T cells resulted in M-CSF secretion of inhibit OC formation (data not shown), this shows that
Ϯ 0.5 ng/ml/10 cells in a 48-h period. No M-CSF biologically active RANKL is secreted by T cells into the
6
secretion was detected in unstimulated T cells.
medium and plays an important, although nonexclusive,
Using semiquantitative RT-PCR, PHA was found to up- role in the osteoclastogenic activity of activated T cell CM.
regulate RANKL expression (Fig. 6A). No RANKL expres- In contrast, treatment with anti–M-CSF antibody at a con-
sion was observed by unstimulated T cells. The same results centration that was found to block RANKL and M-CSF–
were observed using T cells isolated from normal human induced OC formation did not decrease OC formation.
peripheral blood rather than from PBSCs (data not shown). These data suggest that T cells secrete soluble factors that
Because our previous data show that substantial osteo- increase the survival and differentiation of OC precursors in
clastogenic activity is present in activated T cell CM, we a manner similar to M-CSF.
conducted a metabolic labeling experiment to identify the
To further explore the ability of T cell CM to support OC
presence of soluble RANKL. The data show (Fig. 6B) the formation via a RANKL and M-CSF–independent mecha-
immunoprecipitation by anti-RANKL antibody, of a 26- nism, we tested the ability of T cell CM to superinduce OC
kDa protein from PHA-stimulated T cell CM. This protein formation in purified monocytes maximally stimulated with
corresponded exactly to the reported size of soluble rhRANKL (200 ng/ml) and rhM-CSF (100 ng/ml). These
(
13,18)
RANKL.
Secreted RANKL was not detected in un- concentrations of cytokines were determined to be saturat-
ing by dose-response experiments (data not shown). The
stimulated T cell CM.

T-CELLS INDUCE OSTEOCLASTOGENESIS
333
ϩ
ϩ
FIG. 5. Addition of either CD4 or CD8 T cell subsets or CM
FIG. 3. T cell depletion of PBSCs inhibits PHA-induced osteoclas-
togenesis. T-cells were immunomagnetically depleted from PBSC cul-
tures by positive selection using CD3 coated immunomagnetic beads
induces OC formation from monocytes after PHA stimulation. (A)
ϩ
ϩ
PHA-stimulated and -unstimulated CD4 or CD8 T cell subsets,
isolated by antibody coated immunomagnetic beads (see Materials and
Methods section) were added back to purified monocytes. Monocytes
(see Materials and Methods section). Negatively isolated unstimulated
ϩ
ϩ
T cells were replaced in some cultures at the same original concentra- incubated with both CD4 or CD8 T cell subsets stimulated with
tion. In the absence of T cells, the ability of PHA to stimulate OC PHA generated large numbers of OCs. Unstimulated T cells subsets
formation is reduced by 3-fold. Each data point represents the aver- formed only low numbers of OCs. (B) CM from PHA-stimulated
age Ϯ SD of triplicate wells. Similar results were obtained in three CD4 or CD8 T cells (see Materials and Methods section) elicited
independent experiments. p Ͻ 0.01 compared with PHA-stimulated large numbers of OCs from purified monocytes. Each data point rep-
ϩ
ϩ
PBSCs using unpaired two-tailed Student’s t-test.
resents the average Ϯ SD of triplicate wells. Similar results were
obtained in three independent experiments.
DISCUSSION
In this study we have examined the ability of activated T
cells to support human osteoclastogenesis and investigated
the cell populations and cytokines involved, using a well-
defined stromal cell free osteoclastogenic culture system,
(16)
which uses PBSCs as a source of OC precursors.
PBSC
is a peripheral blood product enriched 2- to 20-fold in
ϩ
CD34 stem cells. Because of the enriched number of OC
FIG. 4. Addition of purified T cells to purified monocytes induces
OC formation in the presence of PHA. Human monocytes were isolated
from PBSCs as described in the Materials and Methods section. Puri-
fied T cells isolated by negative selection were added to some wells and
the cultures stimulated with PHA (1.5 ␮g/ml). In the presence of
PHA-stimulated T cells, monocytes generate numerous OCs. Neither
unstimulated nor PHA-stimulated isolated T cells or monocytes gen-
erated OCs independently. Each data point represents the average Ϯ
precursors, these cultures generate relatively large numbers
of OCs compared with culture systems using nonmobilized
blood. However, the number of OCs formed varies between
experiments depending on the degree of CD34 enrichment
achieved in individual donors.
Our data show that activated T cells are indeed capable of
stimulating the production of authentic OCs from human
SD of triplicate wells. Similar results were obtained in three indepen- PBSCs, consistent with previous data from animal mod-
(7,8,10)
dent experiments
els.
We show that the effect of T cells on osteoclas-
togenesis is directed on monocytes (a source of early OC
precursors) and that no other accessory cells are required for
data show (Fig. 8) that activated T cell CM was capable of this effect. Previous studies have suggested that PHA may
eliciting a further ϳ30% increase in OC formation over and have direct effects on OC precursors because PHA-
above that induced by saturating concentrations of RANKL stimulated human mononuclear cells increased bone resorp-
(8)
and M-CSF. This data confirm the ability of T cell CM to tion in fetal rat bone organ cultures. This effect may have
support osteoclastogenesis through RANKL and M-CSF– been caused by a carryover of PHA because our data show
independent mechanisms.
no tendency of PHA to induce OC formation from mono-

3
34
WEITZMANN ET AL.
FIG. 7. The RANKL inhibitor OPG blunts OC formation stimulated
by PHA-activated T cell CM (generated as described in the Materials
and Methods section). Purified monocytes from PBSCs were induced to
form OCs by stimulation with one-quarter volume (final) of PHA-
stimulated T cell CM in the presence or absence of the RANKL
inhibitor OPG (1 ␮g/ml) and/or neutralizing anti–M-CSF antibody
(20␮g/ml). Data representative of three independent experiments. p Ͻ
0
.01 with respect to unstimulated T cell CM as accessed by one-way
ANOVA and Fisher protected LSD test.
OC precursors present in PBSCs versus unmobilized blood
may provide a greater opportunity for the rapid induction of
OCs compared with other systems such as that of Horwood
FIG. 6. PHA up-regulates RANKL in T cells. (A) Negatively isolated
T cells from PBSCs were stimulated with PHA for 24 h and RNA
isolated. Semiquantitative RT-PCR was conducted as described in the et al.
(10)
Materials and Methods section. PHA-stimulated T cells showed ex-
pression of RANKL. No RANKL message was detectable in unstimu- mation in PBSC cultures, some OC formation still was
lated T cells. (B) Secretion of RANKL into the culture supernatant was
measured by metabolic labeling of purified T cells followed by
RANKL immunoprecipitation and SDS-PAGE (see Materials and
Methods section). Soluble RANKL was detected in PHA-activated T
cells but not detected in unstimulated cells. Similar results were ob-
tained in three independent experiments.
Although T cell depletion dramatically reduced OC for-
observed in the presence of PHA. This may be caused by the
Ϫ
presence of T cell precursors (CD3 T cells) in PBSCs,
which mature during the culture period and support low
levels of OC formation.
ϩ
ϩ
Our data show that both CD4 and CD8 T cell subsets
are equally capable of stimulating OC formation and are
(5)
consistent with previous reports in mice and with studies
ϩ
ϩ
showing that both activated CD4 and CD8 T cells ex-
(
20)
cytes in the absence of T cells. We conclude that T cells are press RANKL.
the targets of PHA in this system and the source of osteo- formation
clastogenic activities.
Previous reports of T cells inhibiting OC
(
21)
ϩ
and of CD4 T cells inhibiting OC formation
while CD8 cells stimulate it are in contrast to our data
ϩ
(6)
(9,10)
Although the longevity of T cells would be expected to be and those of other recent studies.
The bulk of the recent
reduced after activation by PHA or CD3 immunomagnetic data are now in favor of the hypothesis that T cells (both
ϩ
ϩ
isolation, we did not find that the addition of fresh T cells to CD4 and CD8 ) are stimulatory to osteoclastogenesis.
the cultures at regular intervals, as has been reported by However, in individual model systems in vitro, many fac-
(10)
other researchers,
was necessary to generate large num- tors including mechanism of T cell activation, differences in
bers of OCs in this system. T cells were found to remain culture conditions, species variations, sources, and prepara-
viable for at least 3 days in culture as assessed by trypan tion of biological material may result in different patterns of
blue dye exclusion. Whether they continue to produce cy- cytokine expression by T cells, leading to apparent conflict-
tokines is unknown; however, the presence of T cells in the ing data. For example, T cells are known to be capable of
cultures with RANKL on their membranes (even if the cells secreting many different cytokines, both osteoclastogenic
are no longer actively secreting cytokines) may account for such as RANKL, tumor necrosis factor ␣ (TNF-␣), and IL-6
the apparent longevity of the response. Where CM was and antiosteoclastogenic such as IL-18, GM-CSF, and in-
used, it was replaced at 3-day intervals thus replenishing terferon ␥ (IFN-␥). Thus, T cells have the capacity to both
spent cytokines. In addition, the large number of potential support or inhibit OC formation.

T-CELLS INDUCE OSTEOCLASTOGENESIS
335
(9)
RANKL. However, in addition to the soluble species of
RANKL and M-CSF detected in our system, membrane
bound forms of M-CSF and RANKL on T cell membranes
also could play important roles in baseline and stimulated
osteoclastogenesis in vivo.
The inability of saturating concentrations of OPG to
block a residual 30% of the osteoclastogenic activity of T
cell CM suggests the presence of additional unidentified
factors capable of inducing OC formation by a RANKL-
independent mechanism. This notion is supported by the
additional observation that T cell CM is able to elicit an
additional 30% OC formation in cultures of monocytes
maximally stimulated with RANKL and M-CSF at concen-
trations found to be saturating. Recently, a number of
RANKL-independent cytokines have been reported to be
(23,24)
capable of inducing OC formation including TNF-␣.
However, in our system recombinant TNF-␣ failed to elicit
any OC formation in T cell–depleted PBSC cultures, over a
wide range of concentration. In addition, treating T cell CM
with neutralizing antibodies to TNF-␣ failed to block either
RANKL-dependent or RANKL-independent OC formation.
This suggests that TNF-␣ neither induces OC formation
independently nor synergizes with RANKL to induce os-
teoclastogenesis in this system (data not shown). We thus
conclude that TNF-␣ is not the cytokine responsible for the
RANKL-independent T cell–mediated osteoclastogenesis in
FIG. 8. T cell CM superinduces OC formation in the presence of
saturating concentrations of RANKL and M-CSF. Purified monocytes
from PBSCs were induced to form OCs by stimulation with saturating
concentrations of RANKL and M-CSF. PHA-stimulated T cell CM
(generated as described in the Materials and Methods section) was
added to some wells and OC counted after 7 days. Each data point
represents the average Ϯ SD of triplicate wells. Similar results were
obtained in three independent experiments. p Ͻ 0.05 with respect to
RANKL and M-CSF–stimulated monocytes alone, using unpaired two- this system. The nature of the RANKL-independent mech-
tailed Student’s t-test.
anism remains unelucidated at this time.
Although the ability of activated T cells to produce
M-CSF messenger RNA (mRNA) and to express the mem-
Of considerable interest is the demonstration of substan- brane bound form of M-CSF has been reported previ-
(25,26)
tial osteoclastogenic activity in PHA-activated T cell CM. ously,
here, for the first time, we show that activated T
The ability of stromal cells to support osteoclastogenesis is cells also produce large concentrations of the soluble form
(
22)
known to require cell-cell contact.
The secretion of os- of M-CSF. Interestingly, despite previous reports that
(12,13)
teoclastogenic factors into the circulation allows activated T M-CSF is an important cofactor for RANKL activity,
cells to not only impact local sites of inflammation by our data show that complete neutralization of M-CSF by
means of membrane bound factors but also to play a crucial anti–M-CSF antibody failed to impact OC survival or block
role in the systemic bone loss characteristic of rheumatoid OC formation in this culture system. Although it is possible
arthritis, through production of secreted factors. The iden- that factors present in T cell CM induce membrane bound
tification of secreted RANKL by PHA-stimulated T cells M-CSF from the macrophages in culture, the high concen-
accounts for some of the soluble osteoclastogenic activity trations of polyclonal antibody used would be expected to
observed in our system. This observation is consistent with block any activity from this potential source. This suggests
other reports of secreted forms of biologically active that activated T cells secrete additional cytokines capable of
(9,18)
RANKL
by activated T cells. These data are also in supporting macrophage survival and OC formation in the
agreement with recent reports of Con A–stimulated T cells absence of M-CSF. One factor that recently has been re-
(10)
supporting osteoclast formation via RANKL expression
ported to be capable of substituting for M-CSF in RANKL-
and the demonstration that in vivo activated T cells play a dependent OC formation is vascular endothelial growth
(27)
key role in the OC formation and bone resorption associated factor (VEGF).
However, combined neutralizing anti-
with rheumatoid arthritis via a RANKL-dependent mecha- bodies against both M-CSF and VEGF were unable to block
(9)
nism.
OC formation, suggesting the presence of other substituting
Because T cell CM alone was able to generate approxi- factors.
mately the same number of OCs as whole T cells, the data
Previous studies in the M-CSF–deficient (op/op) mouse
suggest that soluble factors are dominant in this system. also have indicated that cytokines such as IL-3 and GM-
However, because CM was replenished twice weekly, this CSF can partially compensate for M-CSF deficiency in
(28)
could account for the strong activity of T cell CM. In older mice.
cultures containing whole T cells, the presence of mem-
brane bound factors may have sustained osteoclastogenesis M-CSF–mediated OC formation is activin A.
Another factor recently found to synergize with RANKL/
(29)
Although
in the face of declining cytokine secretion as T cells aged. activin A has not been described as a T cell product, we
Recently, it has been shown that fixed T cells are capable of cannot exclude a potentiating effect of this factor on the
supporting OC formation because of membrane bound RANKL-dependent OC formation in this system. However,

3
36
WEITZMANN ET AL.
because this factor potentiates RANKL-mediated osteoclas- 10. Horwood NJ, Kartsogiannis V, Quinn JM, Romas E, Martin
TJ, Gillespie MT 1999 Activated T lymphocytes support os-
togenesis it is unlikely that activin A plays a role in the
teoclast formation in vitro. Biochem Biophys Res Commun
RANKL-independent OC formation observed.
Although transforming growth factor ␤ (TGF-␤) has been
2
65:144–150.
1
1. Anderson DM, Maraskovsky E, Billingsley WL, Dougall WC,
Tometsko ME, Roux ER, Teepe MC, DuBose RF, Cosman D,
Galibert L 1997 A homologue of the TNF receptor and its
ligand enhance T-cell growth and dendritic-cell function. Na-
ture 390:175–179.
2. Yasuda H, Shima N, Nakagawa N, Yamaguchi K, Kinosaki M,
Mochizuki S, Tomoyasu A, Yano K, Goto M, Murakami A,
Tsuda E, Morinaga T, Higashio K, Udagawa N, Takahashi N,
Suda T 1998 Osteoclast differentiation factor is a ligand for
osteoprotegerin/osteoclastogenesis-inhibitory factor and is
identical to TRANCE/RANKL. Proc Natl Acad Sci USA
95:3597–3602.
3. Lacey DL, Timms E, Tan HL, Kelley MJ, Dunstan CR, Bur-
gess T, Elliott R, Colombero A, Elliott G, Scully S, Hsu H,
Sullivan J, Hawkins N, Davy E, Capparelli C, Eli A, Qian YX,
Kaufman S, Sarosi I, Shalhoub V, Senaldi G, Guo J, Delaney
J, Boyle WJ 1998 Osteoprotegerin ligand is a cytokine that
regulates osteoclast differentiation and activation. Cell 93:
found to act as an important cofactor in OC formation by
(29,30)
RANKL and M-CSF,
in the PBSC system, we have
found that TGF-␤ inhibits OC formation at all concentra-
(31)
tions tested
and thus is not likely to play a role in OC
1
formation in this system.
In conclusion our data support a role for T lymphocytes in
the spontaneous formation of low numbers of OCs observed
in stromal cell free peripheral blood mononuclear cell cul-
tures and previous studies suggesting a critical role for T
cells in inflammatory conditions such as rheumatoid arthri-
tis and periodontitis by a RANKL-mediated pathway. How-
ever, our data also suggest the presence of additional re-
dundant mechanisms for OC formation by T cells that do
not involve RANKL or M-CSF.
1
1
1
65–176.
4. Wong BR, Josien R, Lee SY, Sauter B, Li HL, Steinman RM,
Choi Y 1997 TRANCE (tumor necrosis factor [TNF]-related
activation-induced cytokine), a new TNF family member pre-
dominantly expressed in T cells, is a dendritic cell-specific
survival factor. J Exp Med 186:2075–2080.
ACKNOWLEDGMENTS
This study was supported in part by grants from the
National Institutes of Health (AG 13534 from the National
Institute on Aging [R.P.] and AR-99–001 [M.N.W.], and 15. Faust J, Lacey DL, Hunt P, Burgess TL, Scully S, Van G, Eli
AR46370 [L.R.] from the National Institute of Arthritis and
Musculoskeletal and Skin Disease).
A, Qian Y, Shalhoub V 1999 Osteoclast markers accumulate
on cells developing from human peripheral blood mononuclear
precursors. J Cell Biochem 72:67–80.
1
6. Matayoshi A, Brown C, DiPersio JF, Haug J, Abu-Amer Y,
Liapis H, Kuestner R, Pacifici R 1996 Human blood-mobilized
hematopoietic precursors differentiate into osteoclasts in the
absence of stromal cells. Proc Natl Acad Sci USA 93:10785–
10790.
REFERENCES
1
2
3
4
5
. Duke-Cohan JS, Weinberg H, Sharon R, Naor D 1985 Immu-
nological function in osteoporosis. Clin Immunol Immuno- 17. Harlow E, Lane D 1988 Cell staining. In: Antibodies: A
pathol 35:125–129.
Laboratory Manual. Cold Spring Harbor Laboratories, Cold
Spring Harbor, NY, USA, pp. 359–420.
. Rubin KR, Ballow M, Baron R, Greenstein RM, Raisz LG,
Rowe DW 1988 Malignant osteoporosis and defective immu- 18. Lum L, Wong BR, Josien R, Becherer JD, Erdjument-
noregulation. J Bone Miner Res 3:509–516.
Bromage H, Schlondorff J, Tempst P, Choi Y, Blobel CP
1999 Evidence for a role of a tumor necrosis factor-alpha
(TNF-alpha)-converting enzyme-like protease in shedding of
TRANCE, a TNF family member involved in osteoclastogen-
esis and dendritic cell survival. J Biol Chem 274:13613–
13618.
. Fujita T, Matsui T, Nakao Y, Watanabe S 1984 T lymphocyte
subsets in osteoporosis. Effect of 1-alpha hydroxyvitamin D3.
Miner Electrolyte Metab 10:375–378.
. Ukai T, Hara Y, Kato I 1996 Effects of T cell adoptive transfer
into nude mice on alveolar bone resorption induced by endo-
toxin. J Periodontal Res 31:414–422.
. Hara Y, Ukai T, Yoshimura A, Shiku H, Kato I 1998 His-
topathological study of the role of CD4- and CD8-positive T
cells on bone resorption induced by Escherichia coli endo-
toxin. Calcif Tissue Int 63:63–66.
19. Khalkhali-Ellis Z, Collin-Osdoby P, Li L, Brandi ML, Osdoby
P 1997 A human homolog of the 150 kD avian osteoclast
membrane antigen related to superoxide dismutase and essen-
tial for bone resorption is induced by developmental agents
and opposed by estrogen in FLG 29.1 cells. Calcif Tissue Int
60:187–193.
6
7
8
. John V, Hock JM, Short LL, Glasebrook AL, Galvin RJ 1996
ϩ
A role for CD8 T lymphocytes in osteoclast differentiation in 20. Josien R, Wong BR, Li HL, Steinman RM, Choi Y 1999
vitro. Endocrinology 137:2457–2463.
TRANCE, a TNF family member, is differentially expressed
on T cell subsets and induces cytokine production in dendritic
cells. J Immunol 162:2562–2568.
. Wang TM, Jee WS, Woodbury LA, Matthews JL 1982 Effects
of phytohemagglutinin-P (PHA-P) on bone of the growing rat.
Metab Bone Dis Relat Res 4:193–199.
21. Massey HM, Flanagan AM 1999 Human osteoclasts derive
from CD14-positive monocytes. Br J Haematol 106:167–170.
. Holak HM, Szymaniec S, Skriabin A 1984 Effect of dichlo-
romethylene diphosphonate (Cl2-MDP) on bone resorption 22. Jimi E, Nakamura I, Amano H, Taguchi Y, Tsurukai T,
induced by PHA-stimulated human mononuclear cells in organ
culture. Arch Immunol Ther Exp 32:77–83.
. Kong YY, Feige U, Sarosi I, Bolon B, Tafuri A, Morony S,
Tamura M, Takahashi N, Suda T 1996 Osteoclast function is
activated by osteoblastic cells through a mechanism involving
cell-to-cell contact. Endocrinology 137:2187–2190.
9
Capparelli C, Li J, Elliott R, McCabe S, Wong T, Campag- 23. Azuma Y, Kaji K, Katogi R, Takeshita S, Kudo A 2000 Tumor
nuolo G, Moran E, Bogoch ER, Van G, Nguyen LT, Ohashi
PS, Lacey DL, Fish E, Boyle WJ, Penninger JM 1999 Acti-
necrosis factor-alpha induces differentiation of and bone re-
sorption by osteoclasts. J Biol Chem 275:4858–4864.
vated T cells regulate bone loss and joint destruction in adju- 24. Kobayashi K, Takahashi N, Jimi E, Udagawa N, Takami M,
vant arthritis through osteoprotegerin ligand. Nature 402:304–
09.
Kotake S, Nakagawa N, Kinosaki M, Yamaguchi K, Shima N,
Yasuda H, Morinaga T, Higashio K, Martin TJ, Suda T 2000
3

T-CELLS INDUCE OSTEOCLASTOGENESIS
337
Tumor necrosis factor alpha stimulates osteoclast differentia- 29. Fuller K, Bayley KE, Chambers TJ 2000 Activin A is an
tion by a mechanism independent of the ODF/RANKL-RANK
essential cofactor for osteoclast induction. Biochem Biophys
interaction. J Exp Med 191:275–286.
Res Commun 268:2–7.
2
2
5. Zisman E, Waisman A, Ben-Yair E, Tartakovsky B 1993 30. Sells Galvin RJ, Gatlin CL, Horn JW, Fuson TR 1999 TGF-
Production of colony-stimulating factor 1 by T cells: Possible
involvement in their interaction with antigen-presenting cells.
Cytokine 5:309–318.
beta enhances osteoclast differentiation in hematopoietic cell
cultures stimulated with RANKL and M-CSF. Biochem Bio-
phys Res Commun 265:233–239.
6. Cerdan C, Razanajaona D, Martin Y, Courcoul M, Pavon C, 31. Weitzmann MN, Cenci S, Haug J, Brown C, DiPersio J,
Mawas C, Olive D, Birg F 1992 Contributions of the CD2 and
CD28 T lymphocyte activation pathways to the regulation of
the expression of the colony-stimulating factor (CSF- 1) gene.
J Immunol 149:373–379.
Pacifici R 2000 B lymphocytes inhibit human osteoclastogen-
esis by secretion of TGFbeta. J Cell Biochem 78:318–324.
Address reprint requests to:
Roberto Pacifici, M.D.
Division of Bone and Mineral Diseases
Barnes-Jewish Hospital-North
216 South Kingshighway Boulevard
St. Louis, MO 63110, USA
2
7. Niida S, Kaku M, Amano H, Yoshida H, Kataoka H, Nish-
ikawa S, Tanne K, Maeda N, Kodama H 1999 Vascular
endothelial growth factor can substitute for macrophage
colony-stimulating factor in the support of osteoclastic bone
resorption. J Exp Med 190:293–298.
8. Myint YY, Miyakawa K, Naito M, Shultz LD, Oike Y,
Yamamura K, Takahashi K 1999 Granulocyte/macrophage
colony-stimulating factor and interleukin-3 correct osteopetro-
2
sis in mice with osteopetrosis mutation. Am J Pathol 154:553– Received in original form January 29, 2000; in revised form July
66. 10, 2000; accepted September 1, 2000.
5