Pathological role of osteoclast costimulationin arthritis-induced bone lossSae Ochi*†, Masahiro Shinohara*, Kojiro Sato*‡, Hans-Jurgen Gober*, Takako Koga*, Tatsuhiko Kodama§,Toshiyuki Takai¶, Nobuyuki Miyasaka†, and Hiroshi Takayanagi*�**

Departments of *Cell Signaling and †Medicine and Rheumatology, Graduate School, Tokyo Medical and Dental University, Center of Excellence Program forFrontier Research on Molecular Destruction and Reconstruction of Tooth and Bone, Yushima 1-5-45, Bunkyo-ku, Tokyo 113-8549, Japan; ‡Division ofRheumatology and Applied Immunology, Department of Medicine, Faculty of Medicine, Saitama Medical University, Morohongo 38, Moroyama, Iruma-gun,Saitama 350-0495, Japan; §Department of Molecular Biology and Medicine, Research Center for Advanced Science and Technology, University of Tokyo,Komaba 4-6-1, Meguro-ku, Tokyo 153-8904, Japan; ¶Department of Experimental Immunology, Institute of Development, Aging and Cancer, TohokuUniversity, Seiryo-cho 4-1, Aoba-ku Sendai, Miyagi 980-8575, Japan; and �Solution Oriented Research for Science and Technology, Japan Science andTechnology Agency, Honcho 4-1-8, Kawaguchi-shi, Saitama 332-0012, Japan

Edited by Tak Wah Mak, University of Toronto, Toronto, ON, Canada, and approved May 29, 2007 (received for review March 5, 2007)

Abnormal T cell immune responses induce aberrant expression ofinflammatory cytokines such as TNF-�, leading to osteoclast-mediated bone erosion and osteoporosis in autoimmune arthritis.However, the mechanism underlying enhanced osteoclastogenesisin arthritis is not completely understood. Here we show that TNF-�contributes to inflammatory bone loss by enhancing the osteoclas-togenic potential of osteoclast precursor cells through inducingpaired Ig-like receptor-A (PIR-A), a costimulatory receptor forreceptor activator of NF-�B (RANK). In fact, bone erosion andosteoporosis, but not inflammation, caused by aberrant TNF-�expression were ameliorated in mice deficient in Fc receptorcommon � subunit or �2-microglobulin, in which the expression ofPIR-As and PIR-A ligands is impaired, respectively. These resultsestablish the pathological role of costimulatory receptors for RANKin bone loss in arthritis and may provide a molecular basis for thefuture therapy of inflammatory diseases.

paired immunoglobulin-like receptor � rheumatoid arthritis � TNF-� �costimulatory

Osteoclasts, multinucleated cells of hematopoietic origin thatdegrade the bone matrix (1, 2), are regulated by immuno-

regulatory molecules under both physiological and pathologicalconditions (3). Osteoclast costimulation is an emerging conceptthat has recently come into acceptance based on the observationthat combined deficiency of Fc receptor common � subunit (FcR�)and DNAX-activation protein 12 (DAP12) results in a completelack of osteoclasts (4, 5). In addition to receptor activator of NF-�B(RANK), the receptor for RANK ligand (RANKL), the Ig-likereceptors associated with FcR� and DAP12 have been recognizedas essential receptors for osteoclastogenesis (2, 3). Ig-like receptorswere extensively studied in natural killer and myeloid cells, as wellas in B cells, but this observation established that Ig-like receptorsfunction as osteoclast costimulatory receptors, which are crucial forbone homeostasis under physiological conditions. The mutation inDAP12 causes Nasu-Hakola disease in humans (6, 7), but there hasbeen no other report on the contribution of osteoclast costimulationin pathological conditions in the skeletal system.

Ig-like receptors associate with adaptors harboring an immuno-receptor tyrosine-based activation motif such as FcR� and DAP12in immune cells (6). In osteoclasts, Ig-like receptors that associatewith FcR� include paired Ig-like receptor-A (PIR-A) and oste-oclast-associated receptor (OSCAR), whereas those that associatewith DAP12 include triggering receptor expressed on myeloidcells-2 (TREM-2) and signal-regulatory protein � (SIRP�1) (4).Although the ligands for these receptors have not been wellcharacterized in the context of osteoclastogenesis, it is likely that theligands for DAP12-associating receptors are expressed in theculture system of osteoclast precursor cells, but those for FcR�-associating receptors are mainly expressed by osteoclastogenesis-supporting cells (4, 8). RANK and its costimulatory receptors

synergistically initiate the activation of calcium signaling that leadsto the induction and autoamplification of nuclear factor of activatedT cells c1 (NFATc1) (4), the key transcription factor for osteoclas-togenesis (9, 10).

Rheumatoid arthritis (RA) is an autoimmune disease char-acterized by inflammation of synovial joints with CD4� T cellinfiltration and synovial cell proliferation, leading to severe bonedestruction mediated by osteoclasts (11). Bone loss is not onlyobserved as erosion in the affected joints, but also in the formsof periarticular and systemic osteoporoses (12). Importantly,inflammatory cytokines directly act on osteoclast precursor cellsof hematopoietic lineage and activate their differentiation intoosteoclasts by cooperating with RANKL (13–15), but it is notfully understood how these cytokines exert their direct effect.Although the higher circulation level of these cytokines maycontribute to enhanced osteoclastogenesis in bone tissues apartfrom inflammatory lesions, the mechanism of systemic osteo-porosis associated with arthritis also remains unclear.

TNF-� is one of the critical cytokines in the pathogenesis ofRA, as shown by many gain- and loss-of-function genetic models(16, 17), as well as by the clinical efficacy of anti-TNF-� therapy(18). It is notable that transgenic mice that express humanTNF-� (hTNFtg mice) spontaneously develop destructive arthri-tis (16). Interestingly, the anti-TNF-� antibody has been shownto suppress bone damage in patients who had no clinicalimprovement in terms of pain and inflammation (19), suggestingthat, under arthritic condition, TNF-� exerts an important directaction on bone, which is independent of its action on the immunesystem. Although there has been no in vivo evidence that TNF-�induces osteoclastogenesis in mice lacking RANKL signaling orthat TNF-� rescues osteopetrosis in such mice (20–22), TNF-�clearly acts on osteoclast precursor cells and changes the re-sponsiveness of the cells under certain conditions (13–15).However, the molecular mechanism underlying the enhanced

Author contributions: S.O. performed research; S.O., M.S., T. Koga, K.S., H.-J.G., T. Kodama,T.T., and N.M. analyzed data; T. Kodama and T.T. provided new reagents/analytic tools; andH.T. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Abbreviations: �2M, �2-microglobulin; BMD, bone mineral density; BMM, bone marrow-derived monocyte/macrophage lineage cell; BV/TV, bone volume per tissue volume; DAP12,DNAX-activation protein 12; FcR�, Fc receptor common � subunit; NFATc1, nuclear factorof activated T cells c1; OSCAR, osteoclast-associated receptor; PIR, paired Ig-like receptor;RA, rheumatoid arthritis; RANKL, receptor activator of nuclear factor-�B ligand; TRAP,tartrate-resistant acid phosphatase; TRAP� MNCs, multinucleated cells positive for TRAP;TREM-2, triggering receptor expressed on myeloid cells-2; SIRP�1, signal-regulatory protein�1.

**To whom correspondence should be addressed. E-mail: taka.csi@tmd.ac.jp.

This article contains supporting information online at www.pnas.org/cgi/content/full/0701971104/DC1.

© 2007 by The National Academy of Sciences of the USA

11394–11399 � PNAS � July 3, 2007 � vol. 104 � no. 27 www.pnas.org�cgi�doi�10.1073�pnas.0701971104

Dow

nloa

ded

by g

uest

on

Dec

embe

r 5,

202

1

osteoclastogenic potential of osteoclast precursor cells in thepresence of TNF-� remains to be elucidated.

Here we examined the effect of TNF-� on osteoclastogenesis ina culture of osteoclast precursor cells stimulated with RANKL andM-CSF. We found that the promotive effect of TNF-� is notablyobserved in the late phase of differentiation characterized byNFATc1 autoamplification. Because NFATc1 autoamplificationdepends on calcium signaling, we explored the involvement ofIg-like receptors that initiate calcium signaling in the effect ofTNF-�. We show that TNF-� specifically induces Ig-like receptorsPIR-As and their ligands, MHC class I molecules (23, 24). The invitro effect of TNF-� on osteoclastogenesis was significantly sup-pressed by PIR-A-Fc addition and abrogated in FcR�-deficientcells. The effect was also markedly suppressed in cells deficient in�2M, an essential subunit for MHC class I molecules (25). Wefurther demonstrate the importance of the TNF-�-mediated in-duction of PIR-As by using hTNFtg mice crossed with mice deficientin FcR� (FcR��/�) or �2M (B2m�/�). Thus, TNF-� rendersosteoclast precursor cells prone to differentiating into osteoclasts byup-regulating costimulation through PIR-As, which contribute toboth local and systemic osteoporoses in arthritis.

ResultsTNF-� Enhances Late Phase of Osteoclast Differentiation and NFATc1Autoamplification. To gain insight into the molecular mechanismunderlying the promotive effect of TNF-� on osteoclastogen-esis, TNF-� was added at various time points to the osteoclastformation system: Bone marrow-derived monocyte/macro-phage lineage cells (BMMs) were stimulated with RANKL andM-CSF, and multinucleated cells positive for tartrate-resistantacid phosphatase (TRAP� MNCs) were counted [supportinginformation (SI) Fig. 7A]. When TNF-� was added to BMMsat the same time as RANKL or before it, TNF-� showed nopromotive effect on osteoclastogenesis (Fig. 1A), consistentwith a previous report (13). The promotive effect of TNF-� onTRAP� MNC formation was prominently observed whenTNF-� was added 1 day after RANKL stimulation or later(Fig. 1 A and SI Fig. 7B), and such an effect was more distinctlyobserved at lower RANKL concentrations (Fig. 1B). Theseresults suggest that the effect of TNF-� is not dependent onRANK or signaling molecules immediately activated byRANKL stimulation, such as NF-�B or TNF receptor-associated factor 6 (2). Thus, we evaluated the effect of TNF-�on RANKL-mediated NFATc1 induction, a hallmark event inthe late phase of osteoclastogenesis (9). Both NFATc1 mRNAand protein levels were highly up-regulated by TNF-� (Fig. 1

C and D), suggesting that TNF-� has an effect on moleculesregulating NFATc1 autoamplification.

TNF-� Promotes Osteoclastogenesis Through Induction of PIR-As andTheir Ligands. NFATc1 autoamplification requires calcium signal-ing, which is mediated by costimulatory receptors for RANK andimmunoreceptor tyrosine-based activation motif-harboring adap-tors (4, 8). GeneChip analysis showed that, among Ig-like receptorsthat function as costimulatory receptors, Pira expression was se-lectively increased by TNF-� stimulation in BMMs (Fig. 2A)together with a 2-fold increase in the expression level of FcR� (datanot shown). There are at least six Pira genes in the mouse chro-mosome 7 that encode PIR-A molecules containing a similarectodomain with six Ig-like loops and a transmembrane domainpossessing a positively charged arginine critical for the associationwith FcR� (24). TNF-�-mediated induction of PIR-As was alsoconfirmed by real-time PCR and immunoblot analyses (Fig. 2B andSI Fig. 8).

PIR-A-mediated signals are counterbalanced by the relatedinhibitory receptor PIR-B (26). Although TNF-� also increased thePIR-B expression level, the ratio of the total protein level of PIR-Asto that of PIR-B was markedly increased by TNF-� (Fig. 2B),indicating that the PIR-A-mediated signal is strengthened afterTNF-� stimulation. PIR-A ligands have not been fully determined,but recent reports indicate that they include MHC class I molecules(23, 24). In fact, the expression levels of MHC class I (H-2)molecules such as H-2Db and H-2Kb in BMMs were significantlyincreased by TNF-� (Fig. 2C).

As mentioned earlier, PIR-A ligands are mainly expressed byosteoblasts, and the PIR-A signal is not activated in the culture ofBMMs under physiological conditions (4, 8). However, the TNF-�-mediated induction of PIR-As and their ligands may coopera-tively strengthen the PIR-A signaling axis in the BMM culture andmay contribute to the enhanced osteoclastogenesis under inflam-matory conditions. Consistent with this notion, PIR-A-Fc fusionprotein, but not control IgG, significantly inhibited a TNF-�-mediated enhancement of osteoclast formation from BMMs(Fig. 2D).

A B

C D

Fig. 1. Promotive effects of TNF-� on RANKL-induced osteoclastogenesis andNFATc1 autoamplification. (A) Effect of TNF-� on osteoclastogenesis. Number ofTRAP� MNCs (�3 or �20 nuclei) was counted. See SI Fig. 7A for the treatmentperiods. *, P � 0.05; **, P � 0.01. (B) Effect of TNF-� on osteoclastogenesis in thepresence of RANKL (1 or 5 ng/ml). BMMs cultured in the presence of RANKL (1 or5 ng/ml) were stimulated with TNF-�. *, P � 0.05; **, P � 0.01. (C) Expression ofNfatc1 mRNA in TNF-�-stimulated BMMs (RT-PCR analysis). (D) Expression ofNFATc1 in TNF-�-stimulated BMMs (immunoblot analysis).

Fold

incr

ease

in T

RA

PM

NC

num

ber

0

1

2

3

4

5

TNF-Control

IgGPIR-A

-Fc

RANKL 1 ng/ml

*

Control IgG

PIR-A-Fc

RANKL 5 ng/ml

*

0

1

2

3

Fol

d in

crea

se

Pira1

Pira4Pira

6

Oscar

Sirpb1

Trem

20

2

4

Pirb

TNF-TNF-

PIR-A

SIRP 1

OSCAR

TREM-2

-actin

TNF-50

RANKL(ng/ml) 50

PIR-B

Fol

d in

crea

se in

mR

NA TNF-

TNF-

0

2

4

6

H-2Db

H-2Kb

Rel

ativ

e ra

tio

PIR-A

OSCAR

SIRP

1

TREM

-2

PIR-B

TNF-TNF-

RANKL 5 ng/ml

0

1

2

3

4

5

C

A B

D

Fig. 2. Selective induction of PIR-As and their ligands in response to TNF-�stimulation. (A) GeneChip analysis of mRNA expression of Ig-like receptors inBMMs cultured with or without TNF-�. (B) Expression of PIR-As in TNF-�-stimulated BMMs (immunoblot analysis, Left). The relative ratio of the proteinlevel in TNF-�-treated cells to that in nontreated cells was calculated (Right).(C) Expression of H-2 mRNAs in TNF-�-stimulated BMMs (real-time PCR anal-ysis). (D) Effect of PIR-A-Fc on osteoclastogenesis in TNF-�-stimulated BMMs.BMMs cultured in the presence of RANKL (1 or 5 ng/ml) and TNF-� weretreated with control IgG or PIR-A-Fc. *, P � 0.05.

Ochi et al. PNAS � July 3, 2007 � vol. 104 � no. 27 � 11395

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Dec

embe

r 5,

202

1

Requirement of FcR� in TNF-�-Mediated Promotion of Osteoclasto-genesis. PIR-A has only a short cytoplasmic tail and requires theassociation with FcR� for cell surface expression and signal trans-duction (26). As expected, PIR-As expression was severely im-paired on the membrane of FcR��/� BMMs (Fig. 3A). BecausePIR-A molecules are encoded by multiple genes and it is hardlypossible to disrupt them all genetically, FcR��/� mice serve as analternative tool for analyzing the loss of function of Pira genesdespite a possible contribution of other associating receptors.

When TNF-� was added to FcR��/� BMMs, the promotiveeffect of TNF-� on osteoclastogenesis was markedly suppressedcompared with that on WT BMMs (Fig. 3B). These results indicatethat TNF-� promotes osteoclastogenesis through FcR�-associatingreceptors, including PIR-As. DAP12�/� BMMs are unable todifferentiate into osteoclasts in the pure BMM culture, but candifferentiate in the coculture with osteoblasts (which may expressligands for FcR�-associating receptors) (4, 8). Interestingly,DAP12�/� BMMs could differentiate into osteoclasts with bone-resorbing activity in BMM culture if stimulated with TNF-� inaddition to RANKL (Fig. 3C and SI Fig. 9). This result furthersupports the notion that FcR�-associating receptors such as PIR-Afunction in the BMM culture under inflammatory conditions. Inaddition, no TNF-�-mediated promotion of osteoclastogenesis wasobserved in BMMs deficient in �2M, which forms the invariablelight chain subunit of MHC class I molecules (25) (which representPIR-A ligands) (Fig. 3D). Taken together, these results suggest theimportance of PIR-A among Ig-like receptors in the TNF-�-mediated activation of osteoclastogenesis.

FcR�-Dependent Enhanced Osteoclastogenic Potential of BMMs fromhTNFtg Mice. hTNFtg mice develop destructive arthritis similar toRA and thus are well suited for the analyses of TNF-�-mediatedpathological processes in autoimmune arthritis. In this model,the aberrant expression of TNF-� causes an abnormal prolifer-ation of the synovium and local inflammation, which lead toosteoclast-mediated local bone erosion and osteoporosis (12).

We isolated BMMs from hTNFtg mice and analyzed them byosteoclast formation assay. Interestingly, the BMMs derivedfrom the hTNFtg mice (hTNFtg BMMs) differentiated intoosteoclasts more efficiently than WT BMMs in response toRANKL (Fig. 4A). Although the mechanisms underlying the

enhanced osteoclastogenesis in hTNFtg mice may include theinduction of RANKL in mesenchymal cells (11, 14, 27) or theincrease in the number of osteoclast precursor cells in vivo (21,28), this result suggests a novel mechanism (i.e., the enhance-ment of the osteoclastogenic potential in the hTNFtg BMMs).Flow-cytometric analysis indicated that freshly isolated hTNFtgBMMs expressed a higher level of PIRs than WT BMMs,supporting the notion that the osteoclastogenic potential in thehTNFtg BMMs is enhanced (Fig. 4B). Interestingly, RT-PCRanalysis revealed the expression of TNF in cultured BMMs to beundetected initially, but induced by RANKL stimulation (SI Fig.10), suggesting that the effect of TNF-� is potentiated during thecourse of osteoclastogenesis in hTNFtg BMMs.

NFATc1 induction by RANKL was more significant in thehTNFtg BMMs than in the WT BMMs (Fig. 4C). Therefore, weexamined the expression of Ig-like receptors under the sameconditions. Consistent with the observation in TNF-�-stimulatedBMMs, the PIR-As’ expression level was higher in the hTNFtgBMMs than in the WT BMMs, which was more obvious afterRANKL stimulation (Fig. 4C). To obtain evidence of the impor-tance of PIR-As in the enhanced osteoclastogenesis in the hTNFtgBMMs, we generated hTNFtg/FcR��/� mice in which PIR-A sig-naling is abrogated. Osteoclast formation and the NFATc1 induc-tion by RANKL in the hTNFtg/FcR��/� BMMs were not soenhanced as those in the hTNFtg BMMs (Fig. 4 D and E), suggestingthat enhanced osteoclastogenesis in hTNFtg mice depends on FcR�and possibly on PIR-As.

Osteoclast Formation, but Not Joint Inflammation, Is Dependent onFcR� in hTNFtg Mice. Because the onset of antigen-induced arthritisis triggered by antibody recognition through Fc receptors, Fc �receptor III and its adaptor subunit, FcR�, are essential for theinduction of antibody-mediated arthritis (29, 30), and the diseaseseverity is enhanced in mice lacking an inhibitory Fc � receptor IIB(31). However, adaptive immune reactions, including antibodyproduction, are not involved in arthritis onset in hTNFtg mice (22).Therefore, we can analyze the role of FcR� in the phase of bonedestruction by crossing hTNFtg mice with FcR��/� mice. As ex-pected, the onset and clinical course of arthritis in hTNFtg/FcR��/�

mice were comparable to those in hTNFtg mice (data not shown),and there was no marked difference in inflammation severitybetween these mice (Fig. 4 F–H). However, bone destruction andosteoclast formation at erosive lesions in the hTNFtg/FcR��/� micewere markedly suppressed compared with those in hTNFtg mice(Fig. 4 F, I, and J).

As reported previously (12, 32), in addition to bone erosion inthe affected joints, hTNFtg mice presented trabecular bone lossin the subchondral region (periarticular osteoporosis), accom-panied by an increase in osteoclast number (Fig. 4 K–M).Moreover, 20-week-old hTNFtg mice showed systemic osteopo-rosis characterized by a decrease in bone mineral density (BMD)throughout the long bone (Fig. 4N). The decrease in periartic-ular trabecular bone volume [bone volume per tissue volume(BV/TV)] and the increase in osteoclast number per bonesurface, as well as the reduction in BMD, in hTNFtg/FcR��/�

mice were all attenuated compared with those in hTNFtg mice(Fig. 4 K–N). These results suggest that both types of inflam-mation-induced bone loss, namely, local bone erosion andosteoporosis, are dependent on enhanced osteoclast formationthrough FcR�-associating receptors such as PIR-As.

�2M Is Required for Osteoclast Formation, but Not for Joint Inflam-mation, in hTNFtg Mice. To further provide in vivo evidence thatPIR-A-mediated activated osteoclastogenesis contributes to boneloss in hTNFtg mice, we investigated mice with the B2m�/� back-ground, in which the expression of PIR-A ligands (MHC class Imolecules) is compromised. B2m�/� mice develop normally andexhibit no obvious bone phenotype, suggesting that MHC class I

Cel

l num

ber

PIR100 101 102

200

0

Control IgGWTFcR

100

TNF-RANKL (ng/ml)

DAP12

Num

ber

of T

RA

PM

NC

s (/c

m2 )

n.d.n.d.

5

0

20

40

50 WTTNF-

B2m

RANKL1 ng/ml

Fold

incr

ease

in

TRA

PM

NC

num

ber

*

WT B2m

*

0

1

2

0

5

10

0

1

2*

WT

RANKL 5 ng/ml

FcR

RANKL 1 ng/ml

WTTNF-

FcR

Fold

incr

ease

in

TRA

PM

NC

num

ber

*

0

5

10

RANKL 5 ng/ml

A B

C D

Fig. 3. Central role of PIR-As and their adaptor FcR� in enhanced osteoclasto-genesis induced by TNF-�. (A) Cell surface expression of PIRs on BMMs. Flow-cytometric analysis of PIRs expression in WT (thin line) or FcR��/� (thick line)BMMs.Thethick lineshowsWTBMMsstainedwithcontrol IgG. (B)EffectofTNF-�on osteoclastogenesis in FcR��/� BMMs. WT or FcR��/� BMMs cultured in thepresence of RANKL (1 or 5 ng/ml) were stimulated with TNF-�. *, P � 0.05. (C)Effect of TNF-� on osteoclastogenesis in DAP12�/� BMMs. DAP12�/� BMMscultured in the presence of RANKL (5 or 50 ng/ml) were stimulated with TNF-� (50ng/ml). n.d., not detected. (D) Effect of TNF-� on osteoclastogenesis in B2m�/�

BMMs. WT or B2m�/� BMMs cultured in the presence of RANKL (1 or 5 ng/ml)were stimulated with TNF-�. *, P � 0.05.

11396 � www.pnas.org�cgi�doi�10.1073�pnas.0701971104 Ochi et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 5,

202

1

molecules do not play a crucial role in the regulation of bonemetabolism under physiological conditions (25, 33). There was nosignificant difference in clinical course or inflammation severitybetween hTNFtg and hTNFtg/B2m�/� mice (Fig. 5 A and B).However, the degree of bone erosion and the number of osteoclastsin the affected bone lesions in hTNFtg/B2m�/� mice were markedlysuppressed compared with those in hTNFtg mice (Fig. 5 C and D).The increase in the number of osteoclasts, the reduction in sub-chondral bone volume, and the decrease in BMD in hTNFtg/

B2m�/� mice were greatly attenuated compared with those inhTNFtg mice (Fig. 5 E–H), suggesting that periarticular and sys-temic osteoporoses in hTNFtg mice are dependent on �2M expres-sion. Although the role of �2M may not be limited to PIR-Aactivation, these results in toto suggest the importance of the PIR-Aaxis in enhanced osteoclastogenesis and bone loss in arthritis.

Anti-TNF-� Antibody Normalizes the Accelerated Osteoclastogenesisand Inhibits Bone Loss in Arthritis. Finally, we examined the thera-peutic effect of the anti-human TNF-� antibody (infliximab) onbone loss in hTNFtg mice and the enhanced osteoclastogenicpotential of BMMs from these mice. Inflammatory changes and thelocal bone erosion in hTNFtg mice were ameliorated by the ad-ministration of infliximab (Fig. 6 A–D and SI Fig. 11A) to thesimilar extent as reported previously (34). Furthermore, we ob-served a marked therapeutic effect on both periarticular andsystemic osteoporoses accompanied by a decrease in the number ofosteoclasts in subchondral bone (Fig. 6 E–G and SI Fig. 11B).Consistent with this, the enhanced osteoclastogenesis in hTNFtgBMMs was normalized in the presence of infliximab (Fig. 6H). Itis notable that the addition of infliximab repressed the enhancedinduction of PIR-As and NFATc1 observed in the hTNFtg BMMs(Fig. 6I). These results suggest that blocking TNF-� suppressesosteoclastogenesis by inhibiting the PIR-A induction, and thismechanism may underlie, at least in part, the therapeutic effect ofthe anti-TNF-� antibody on bone loss in arthritis.

DiscussionCytokine-targeted therapy is one of the greatest recent advancesin the treatment of autoimmune diseases, including RA. TNF-�is an inflammatory cytokine that has a profound impact on theinnate and adaptive immunities and was first adopted as a targetof a biological therapy (18). Accumulating evidence indicatesthat the anti-TNF-� antibody has beneficial effects on not onlyinflammation, but also bone destruction in arthritis (19). Al-though it has been shown that osteoclasts are effector cellsessential for bone loss in arthritis (22, 35), it has not been fully

PIR

hTNFtgWT

Cel

l num

ber

1021010

100

50

103

I

N.O

c/B

S (/

100

mm

)

0

400

800* *

WT

hTN

Ftg

hTN

Ftg

/Fc

R

NL

0

5

10

15

Art

hriti

s sc

ore n.s.

WT

hTN

Ftg

hTN

Ftg

/Fc

R

Fol

d in

crea

se in

os

teoc

last

num

ber

* *

0

2

4*

WT

hTN

Ftg

hTN

Ftg

/Fc

R

Paw

thic

knes

s (m

m)

WT

hTN

Ftg

hTN

Ftg

/Fc

R

2

2.5

3

3.5 n.s. J

M

BV

/TV

(%

) *

0

5

10

15*

WT

hTN

Ftg

hTN

Ftg

/Fc

R

BM

D (

mg/

cm)

WThTNFtghTNFtg/FcR

# 1 6 11 16 10

20

30

40

50

K WT hTNFtg hTNFtg/FcR

100 m

F WT hTNFtg hTNFtg/FcR

100 m

0.5 mm

Num

ber

of T

RA

PM

NC

s (/c

m2 )

1 5RANKL(ng/ml)

WThTNFtg

*

*

0

200

400

WT hTNFtg50 50

PIR-A

NFATc1

-actin

OSCAR

TREM-2

SIRP 1

RANKL(ng/ml)

CA D

Num

ber

of T

RA

PM

NC

s (/

cm2 )

*

WT

hTN

Ftg

hTN

Ftg

/Fc

RFcR

0

100

200

300

*

WT

hTN

Ftg

0

5

10

His

tolo

gica

l sco

re

NFATc1

-actin

RANKL (ng/ml) 50

FcR

50

hTNFtg

50

WT

50

hTNFtg/FcR

B

hTN

Ftg

/Fc

R

E

G H

Fig. 4. FcR� mediatesbonedestructionandosteoporosis,butnot inflammationin hTNFtg mice. (A) Osteoclastogenesis of hTNFtg BMMs stimulated with RANKL(1 or 5 ng/ml). *, P � 0.05. (B) Flow-cytometric analysis of PIRs’ expression on theWT (dotted line) and hTNFtg (solid line) BMMs. (C) Expressions of NFATc1 andIg-like receptors in hTNFtg BMMs (immunoblot analysis). (D) RANKL-inducedosteoclastogenesis of WT, hTNFtg, FcR��/�, and hTNFtg/FcR��/� BMMs. *, P �0.05. (E) Expression of NFATc1 in WT, hTNFtg, FcR��/�, and hTNFtg/FcR��/� BMMs(immunoblot analysis). (F) (Upper) Histological analysis of ankle joints (H&E).(Lower)Magnifiedviewoftherectangulararea. (G)PawthicknessofWT,hTNFtg,and hTNFtg/FcR��/� mice. n.s., not significant. (H–J) Arthritis score (H), histolog-ical score (I), and fold increase (J) in the number of osteoclasts in the ankles of WT,hTNFtg, and hTNFtg/FcR��/� mice. n.s., not significant; *, P � 0.05; **, P � 0.01.(K) Histological analysis of proximal tibiae of WT, hTNFtg, and hTNFtg/FcR��/�

mice (TRAP and Toluidine blue staining). (L–N) BV/TV (L), number of osteoclastsper bone surface (N.Oc/BS) (M), and BMD (N) of tibiae of WT, hTNFtg, andhTNFtg/FcR��/� mice. *, P � 0.05.

0

5

10

15

BV

/TV

(%

) **

WT

hTN

Ftg

hTN

Ftg

/B2m

0

400

800

N.O

c/B

S (/

100

mm

)

WT

hTN

Ftg

hTN

Ftg

/B2m

* *

A

2

2.5

3

3.5

Paw

thic

knes

s (m

m)

WT

hTN

Ftg

hTN

Ftg

/B

2m

n.s.

WT hTNFtg hTNFtg/B2m

B

0

5

10

15

Art

hriti

s sc

ore

WT

hTN

Ftg

hTN

Ftg

/B2

m

n.s.

His

tolo

gica

l sco

re

WT

hTN

Ftg

hTN

Ftg

/B2

m

0

5

10

C*

E

H

D

0

2

4

Fol

d in

crea

se in

os

teoc

last

num

ber

WT

* **

100 m

WThTNFtghTNFtg/B2m

BM

D (m

g/cm

)

10

20

30

40

50

6 11 16#1

hTN

Ftg

/B2

m

hTN

Ftg

F G

Fig. 5. Involvement of MHC class I molecules in TNF-�-induced bone loss. (A)PawthicknessofWT,hTNFtg, andhTNFtg/B2m�/� mice.n.s.,notsignificant. (B–D)Arthritis score (B), histological score (C), and fold increase (D) in the number ofosteoclasts in the ankles of WT, hTNFtg, and hTNFtg/B2m�/� mice. n.s., notsignificant; *, P � 0.05; **, P � 0.01. (E) Histological analysis of proximal tibiae ofWT, hTNFtg, and hTNFtg/B2m�/� mice (TRAP and Toluidine blue staining). (F–H)BV/TV (F), number of osteoclasts per bone surface (N.Oc/BS) (G), and BMD (H) oftibiae of WT, hTNFtg, and hTNFtg/B2m�/� mice. *, P � 0.05.

Ochi et al. PNAS � July 3, 2007 � vol. 104 � no. 27 � 11397

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Dec

embe

r 5,

202

1

understood how TNF-� activates osteoclastic bone resorptionand exerts its detrimental effects on bone.

As a possible mechanism for TNF-�-mediated enhancement ofosteoclastogenesis, it has been shown that the population of oste-oclast precursor cells is increased in the peripheral blood of hTNFtgmice (21, 28). However, we observed that more osteoclasts wereformed from hTNFtg BMMs even when the same number of BMMswas cultured. TNF-� stimulates the expression of osteoclast differ-entiation factor RANKL by acting on the osteoclastogenesis-supporting mesenchymal cells (11, 14, 27), but it also acts directlyon osteoclast precursor cells and promotes RANKL-induced oste-oclastogenesis (13, 14). In fact, the direct effect plays a critical roleunder certain conditions in vivo (14). To explain the direct effect ofTNF-� on osteoclast precursor cells at the molecular level, it hasbeen reported that TNF-� up-regulates or activates molecules inproximal RANK signaling such as RANK, TNF receptor-associated factor 6, and NF-�B (13, 36–38). We found that nopromotive effect of TNF-� on osteoclastogenesis is observed in theearly phase of differentiation in which these molecules are acti-vated. Instead, the effect of TNF-� is more prominently observedin the late phase, characterized by NFATc1 autoamplification.NFATc1 is an essential and integral transcription factor for oste-oclastogenesis, and its essential role is determined by its specificgene regulatory mechanism of autoamplification (i.e., NFATc1binds to its own promoter and amplifies its own expression) (9). Thenuclear translocation of NFATc1 is regulated by phosphatasecalcineurin, the activation of which depends on calcium signaling(10). Therefore, we inferred that TNF-� targets molecules thatregulate calcium signaling.

Recently, Ig-like receptors and immunoreceptor tyrosine-basedactivation motif-harboring adaptors such as FcR� and DAP12 haveemerged as critical initiators of calcium signaling in osteoclasto-

genesis (costimulatory signal for RANK) (4, 5). Among the Ig-likereceptors involved in osteoclastogenesis, we found the expression ofPIR-As (and their ligands MHC class I molecules) was selectivelyinduced by TNF-�. The series of experiments using FcR��/� andB2m�/� mice, as well as PIR-A-Fc collectively shows that TNF-�activates the PIR-A signaling axis and enhances the ability ofosteoclast precursor cells to differentiate into osteoclasts, thusestablishing an additional mechanism underlying the effect ofTNF-� on bone loss in arthritis. Although the increase in osteoclastnumber was markedly suppressed in hTNFtg/FcR��/� or hTNFtg/B2m�/� mice compared with hTNFtg mice, the BMD in hTNFtg/FcR��/� mice or hTNFtg/B2m�/� was not completely normalized.Further studies are necessary to clearly explain this discrepancy, butit is possible that decreased BMD results from the TNF-�-mediateddecrease in bone formation (39, 40), which was not cancelled evenin the absence of FcR� or �2M. Interestingly, the effect of TNF-�is antiosteoclastogenic if BMMs are exposed to TNF-� beforeRANKL possibly because TNF-�-stimulated BMMs are prone tocommit themselves to activated macrophages (13, 41). It is likelythat this mechanism is inhibited by unknown mechanism(s) in vivobecause we observed that the hTNFtg BMMs have a higherosteoclastogenic potential. Consistent with this, the inhibitoryeffect of TNF-� is less observed in the coculture system of BMMsand osteoblasts (data not shown), suggesting that bone marrowmicroenvironments prevent the BMMs from differentiating intoactivated macrophages even in the abundance of TNF-�.

�2M is a crucial subunit for MHC class I molecules and isessential for the positive selection of CD8� T cells (25). This studyshowed that it plays an important role in the TNF-�-mediatedacceleration of osteoclastogenesis, but �2M is also involved in otherpathological conditions in bones and joints by different mecha-nisms. For example, �2M is related to dialysis-associated amyloidosteoarthropathy (42). In addition, B2m�/� mice crossed withHLA-B27 transgenic mice or B2m�/� mice of susceptible back-grounds develop arthritis spontaneously (43, 44). Therefore, �2M isrequired for suppressing the onset of autoimmune diseases undercertain conditions. However, because the onset of arthritis inhTNFtg mice in the B2m�/� background was not altered in thisstudy, �2M is not involved in the onset of this type of arthritis andexclusively functions in the bone-destruction phase in this model.

A defect in osteoclast costimulation (caused by loss-of-function mutation in DAP12 or TREM-2) is associated with anautosomal recessive condition with bone cysts and preseniledementia called Nasu-Hakola disease (6, 7, 45). Another oste-oclast costimulatory receptor system, PIR, is involved in graft-versus-host diseases (23), but has never been linked to bonediseases. This study reports on the role of the costimulatoryreceptor for RANK in the context of pathological activation ofosteoclastogenesis in inflammation-related bone diseases. Notethat the expression level of human orthologues for PIRs, leu-kocyte Ig-like receptors, in synovial cells in RA is increased (46),but further studies are necessary to determine the functionalsignificance of this finding. Currently, neither PIR-A-Fc appro-priate for the in vivo administration nor the genetically modifiedmice lacking all PIR-A molecules are available, but it will be animportant issue in the future to develop a method to specificallyand completely disrupt PIR-A function in vivo. Because FcR��/�

mice show no obvious defect in the bone homeostasis (4),PIR-As and their adaptor, FcR�, play an important role specif-ically in the pathological activation of osteoclastogenesis. There-fore, targeting the PIR-A signaling axis will be an auspicioustherapeutic strategy in inflammation-induced bone loss.

MethodsIn Vitro Osteoclast Formation. In vitro osteoclast differentiation wasperformed as described previously (47). See SI Methods for details.

BV

/TV

(%

)

N.O

c/B

S (/

100

mm

)

hTN

Ftg

WT

* *

0

200

400

600

E**

0

5

10

15

WThTNFtg

Num

ber

of T

RA

PM

NC

s (/

cm2 ) *

0

100

200H

BM

D (

mg/

cm)

WThTNFtghTNFtg infliximab

6 11 1610

20

30

40

50

#1

hTN

Ftg

inflix

imab

Infliximab

RANKL(ng/ml) 50 50 0

hTNFtgWT

5

NFATc1

I hTNFtginfliximab

*

0

2

4

hTN

Ftg

hTN

Ftg

inflix

imabWTF

old

incr

ease

in ***

2

3

4

hTN

Ftg

hTN

Ftg

inflix

imabWT

*

0

5

10

15*

hTN

Ftg

hTN

Ftg

inflix

imabWT

0

5

10

hTN

Ftg

hTN

Ftg

inflix

imabWT

hTN

Ftg

WT

hTN

Ftg

inflix

imab

A B C D

F G

Fig. 6. Infliximab normalizes enhanced osteoclastogenic potential and boneloss in hTNFtg mice. (A) Paw thickness of WT, hTNFtg, and infliximab-treatedhTNFtg (�infliximab) mice. *, P � 0.05. (B–D) Arthritis score (B), histologicalscore (C), and fold increase in the number of osteoclasts (D) in the ankles of WT,hTNFtg, and infliximab-treated hTNFtg mice. *, P � 0.05; **, P � 0.01. (E–G)BV/TV (E), number of osteoclasts per bone surface (N.Oc/BS) (F) , and BMD (G)of tibiae of WT, hTNFtg, and infliximab-treated hTNFtg mice(hTNFtg�infliximab). *, P � 0.05. (H) Effect of infliximab on osteoclastogen-esis in hTNFtg BMMs. hTNFtg BMMs cultured in the presence of RANKL weretreated with saline (�infliximab) or 100 �g/ml of infliximab (�infliximab).

*, P � 0.05. (I) Effect of infliximab on the expression of NFATc1 and PIR-As inhTNFtg BMMs (immunoblot analysis). Increased expression of PIR-As andNFATc1 in hTNFtg BMMs was inhibited by infliximab (hTNFtg � infliximab).

11398 � www.pnas.org�cgi�doi�10.1073�pnas.0701971104 Ochi et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 5,

202

1

Immunoblot and Flow-Cytometric Analyses. Cell lysates were sub-jected to immunoblot analysis with the following specific anti-bodies: anti-PIR, which recognizes the common domain ofPIR-As and PIR-Bs (band size, 90 kDa and 110 kDa, respec-tively) (48); anti-SIRP�1 (49); anti-OSCAR (R&D Systems,Minneapolis, MN); anti-NFATc1 (Santa Cruz Technologies,Santa Cruz, CA); anti-TREM-2 (R&D Systems); and anti-�-actin (Sigma–Aldrich, St. Louis, MO) antibodies. Expressionlevel was measured by using Scion Image (Scion Corporation,Frederick, MD). After the expression level of costimulatoryreceptors was normalized by that of �-actin, the relative ratio ofthe protein level in TNF-�-treated cells to that in nontreatedcells was calculated in Fig. 2B. For flow-cytometric analysis,BMMs were incubated with the anti-PIR antibody or control ratpolyclonal IgG (BD Biosciences Pharmingen, San Diego, CA)for 30 min, followed by staining with PE-conjugated anti-rat IgG.Cells were analyzed by FACScan using CellQuest software(Becton Dickinson, San Jose, CA).

RT-PCR Analysis. Total RNA was extracted by using Sepazol(Nacalai Tesque, Kyoto, Japan) from BMMs and subjected toRT-PCR analysis. PCR was performed with primers for Pir-b asdescribed previously (50). The sequences of the primers used forNfatc1, GAPDH, Pir-a1, Pir-a4, Pir-a6, TNF, and consensussequences of H-2Db and H-2Kb are described in SI Methods.

GeneChip Analysis. GeneChip analysis was performed as previouslydescribed (10). Briefly, BMMs were stimulated with RANKL (50ng/ml) in the presence of M-CSF for 3 days. TNF-� (50 ng/ml) wasadded 1 day after RANKL stimulation. Total RNA was extractedby Sepazol and used for cDNA synthesis by RT, followed by thesynthesis of biotinylated cRNA through in vitro transcription. AftercRNA fragmentation, hybridization with an MG430v2 GeneChip(Affymetrix, Santa Clara, CA) was performed.

PIR-A-Fc Fusion Protein. cDNA fragments encoding the commonextracellular domain of PIR-As were cloned by using the specific

primers (see SI Methods for details). The extracellular domainwas fused to cDNA for the Fc portion of human IgG1a in frame.PIR-A-Fc fusion protein was prepared as described previously(51). Because the sequences used for the protein are commonlyfound in PIR-A1, PIR-A2, PIR-A3, PIR-A4, PIR-A6, andPIR-A7, this fusion protein inhibits all of these PIR-A isoforms.The fusion protein or control human polyclonal IgG (JacksonImmunoResearch Laboratories, West Grove, PA) was added at0.1 mM on the same day as RANKL stimulation.

Treatment with Infliximab. Sixteen-week-old hTNFtg mice and theirlittermate controls were injected i.p. with 5 mg/kg body weightinfliximab (Tanabe Seiyaku, Osaka, Japan) or saline three times perweek for 4 weeks. Mice were killed and analyzed at the age of 20weeks.

Mice and Clinical/Histological Analyses. See SI Methods.

Statistical Analyses. See SI Methods.

We thank H. Kubagawa, T. Matozaki, L.L. Lanier, and Tanabe Seiyaku Co.,Ltd., for invaluable reagents; and M. Isobe-Ohba, Y. Kim, K. Okamoto, K.Arai, Y. Suzuki, N. Kumazaki, T. Nakashima, M. Asagiri, K. Nishikawa, A.Suematsu, S. Kamano, M. Hayashi, H. Murayama, I. Takayanagi, and K.Takayanagi for helpful discussion and assistance. This work was supported,in part, by a Grant-in-Aid for Creative Scientific Research (to H.T.);Postdoctoral Fellowships for Foreign Researchers from the Japan Societyfor the Promotion of Science (to H.-J.G.); Genome Network Project fromMinistry of Education, Culture, Sports, Science and Technology of Japan(MEXT) grants (to H.T.); grants from MEXT for 21st Century Center ofExcellence (COE) program (to H.T., S.O., and N.M.); the Solution Ori-ented Research for Science and Technology program of the Japan Scienceand Technology Agency (to H.T.); Health Sciences Research Grants fromthe Ministry of Health, Labour, and Welfare of Japan (to H.T.); and grantsfrom Inamori Foundation (to H.T.), Kanae Foundation for Life & Socio-Medical Science (to H.T.), Tokyo Biochemical Research Foundation andYokoyama Foundation for Clinical Pharmacology (to H.T.), NakajimaFoundation (to T. Koga), and Hayashi Memorial Foundation for FemaleNatural Scientists (to T. Koga).

1. Teitelbaum SL, Ross FP (2003) Nat Rev Genet 4:638–649.2. Asagiri M, Takayanagi H (2007) Bone 40:251–264.3. Takayanagi H (2007) Nat Rev Immunol 7:292–304.4. Koga T, Inui M, Inoue K, Kim S, Suematsu A, Kobayashi E, Iwata T, Ohnishi H, Matozaki

T, Kodama T, et al. (2004) Nature 428:758–763.5. Mocsai A, Humphrey MB, Van Ziff le JA, Hu Y, Burghardt A, Spusta SC, Majumdar S,

Lanier LL, Lowell CA, Nakamura MC (2004) Proc Natl Acad Sci USA 101:6158–6163.6. Humphrey MB, Lanier LL, Nakamura MC (2005) Immunol Rev 208:50–65.7. Paloneva J, Kestila M, Wu J, Salminen A, Bohling T, Ruotsalainen V, Hakola P, Bakker AB,

Phillips JH, Pekkarinen P, et al. (2000) Nat Genet 25:357–361.8. Takayanagi H (2005) J Mol Med 83:170–179.9. Asagiri M, Sato K, Usami T, Ochi S, Nishina H, Yoshida H, Morita I, Wagner EF, Mak TW,

Serfling E, Takayanagi H (2005) J Exp Med 202:1261–1269.10. Takayanagi H, Kim S, Koga T, Nishina H, Isshiki M, Yoshida H, Saiura A, Isobe M, Yokochi

T, Inoue J, et al. (2002) Dev Cell 3:889–901.11. Sato K, Takayanagi H (2006) Curr Opin Rheumatol 18:419–426.12. Schett G, Hayer S, Zwerina J, Redlich K, Smolen JS (2005) Nat Clin Pract Rheumatol

1:47–54.13. Lam J, Takeshita S, Barker JE, Kanagawa O, Ross FP, Teitelbaum SL (2000) J Clin Invest

106:1481–1488.14. Kitaura H, Sands MS, Aya K, Zhou P, Hirayama T, Uthgenannt B, Wei S, Takeshita S,

Novack DV, Silva MJ, et al. (2004) J Immunol 173:4838–4846.15. Kim N, Kadono Y, Takami M, Lee J, Lee SH, Okada F, Kim JH, Kobayashi T, Odgren PR,

Nakano H, et al. (2005) J Exp Med 202:589–595.16. Keffer J, Probert L, Cazlaris H, Georgopoulos S, Kaslaris E, Kioussis D, Kollias G (1991)

EMBO J 10:4025–4031.17. Mori L, Iselin S, De Libero G, Lesslauer W (1996) J Immunol 157:3178–3182.18. Feldmann M (2002) Nat Rev Immunol 2:364–371.19. Smolen JS, Han C, Bala M, Maini RN, Kalden JR, van der Heijde D, Breedveld FC, Furst

DE, Lipsky PE (2005) Arthritis Rheum 52:1020–1030.20. Li J, Sarosi I, Yan XQ, Morony S, Capparelli C, Tan HL, McCabe S, Elliott R, Scully S, Van

G, et al. (2000) Proc Natl Acad Sci USA 97:1566–1571.21. Li P, Schwarz EM, O’Keefe RJ, Ma L, Boyce BF, Xing L (2004) J Bone Miner Res 19:207–213.22. Redlich K, Hayer S, Ricci R, David JP, Tohidast-Akrad M, Kollias G, Steiner G, Smolen

JS, Wagner EF, Schett G (2002) J Clin Invest 110:1419–1427.23. Nakamura A, Kobayashi E, Takai T (2004) Nat Immunol 5:623–629.24. Takai T (2005) Immunology 115:433–440.25. Zijlstra M, Bix M, Simister NE, Loring JM, Raulet DH, Jaenisch R (1990) Nature

344:742–746.26. Takai T, Ono M (2001) Immunol Rev 181:215–222.27. Hofbauer LC, Lacey DL, Dunstan CR, Spelsberg TC, Riggs BL, Khosla S (1999) Bone

25:255–259.

28. Li P, Schwarz EM, O’Keefe RJ, Ma L, Looney RJ, Ritchlin CT, Boyce BF, Xing L (2004)Arthritis Rheum 50:265–276.

29. Kleinau S, Martinsson P, Heyman B (2000) J Exp Med 191:1611–1616.30. Ji H, Ohmura K, Mahmood U, Lee DM, Hofhuis FM, Boackle SA, Takahashi K, Holers VM,

Walport M, Gerard C, et al. (2002) Immunity 16:157–168.31. Yuasa T, Kubo S, Yoshino T, Ujike A, Matsumura K, Ono M, Ravetch JV, Takai T (1999)

J Exp Med 189:187–194.32. Herrak P, Gortz B, Hayer S, Redlich K, Reiter E, Gasser J, Bergmeister H, Kollias G,

Smolen JS, Schett G (2004) Arthritis Rheum 50:2044–2047.33. Marusic A, Katavic V, Stimac D, Kusec V, Jonjic S (1995) Eur J Clin Chem Clin Biochem

33:915–918.34. Redlich K, Gortz B, Hayer S, Zwerina J, Doerr N, Kostenuik P, Bergmeister H, Kollias G,

Steiner G, Smolen JS, Schett G (2004) Am J Pathol 164:543–555.35. Pettit AR, Ji H, von Stechow D, Muller R, Goldring SR, Choi Y, Benoist C, Gravallese EM

(2001) Am J Pathol 159:1689–1699.36. Komine M, Kukita A, Kukita T, Ogata Y, Hotokebuchi T, Kohashi O (2001) Bone

28:474–483.37. Kaji K, Katogi R, Azuma Y, Naito A, Inoue JI, Kudo A (2001) J Bone Miner Res 16:1593–1599.38. Abu-Amer Y, Erdmann J, Alexopoulou L, Kollias G, Ross FP, Teitelbaum SL (2000) J Biol

Chem 275:27307–27310.39. Jilka RL, Weinstein RS, Bellido T, Parfitt AM, Manolagas SC (1998) J Bone Miner Res

13:793–802.40. Walsh MC, Kim N, Kadono Y, Rho J, Lee SY, Lorenzo J, Choi Y (2006) Annu Rev Immunol

24:33–63.41. Heidenreich S, Weyers M, Gong JH, Sprenger H, Nain M, Gemsa D (1988) J Immunol

140:1511–1518.42. Bardin T (1987) J Rheumatol 14:647–649.43. Khare SD, Luthra HS, David CS (1995) J Exp Med 182:1153–1158.44. Kingsbury DJ, Mear JP, Witte DP, Taurog JD, Roopenian DC, Colbert RA (2000) Arthritis

Rheum 43:2290–2296.45. Paloneva J, Manninen T, Christman G, Hovanes K, Mandelin J, Adolfsson R, Bianchin M,

Bird T, Miranda R, Salmaggi A, et al. (2002) Am J Hum Genet 71:656–662.46. Tedla N, Gibson K, McNeil HP, Cosman D, Borges L, Arm JP (2002) Am J Pathol

160:425–431.47. Takayanagi H, Ogasawara K, Hida S, Chiba T, Murata S, Sato K, Takaoka A, Yokochi T,

Oda H, Tanaka K, et al. (2000) Nature 408:600–605.48. Kubagawa H, Chen CC, Ho LH, Shimada TS, Gartland L, Mashburn C, Uehara T, Ravetch

JV, Cooper MD (1999) J Exp Med 189:309–318.49. Hayashi A, Ohnishi H, Okazawa H, Nakazawa S, Ikeda H, Motegi S, Aoki N, Kimura S,

Mikuni M, Matozaki T (2004) J Biol Chem 279:29450–29460.50. Kubagawa H, Burrows PD, Cooper MD (1997) Proc Natl Acad Sci USA 94:5261–5266.51. Kim Y, Sato K, Asagiri M, Morita I, Soma K, Takayanagi H (2005) J Biol Chem

280:32905–32913.

Ochi et al. PNAS � July 3, 2007 � vol. 104 � no. 27 � 11399

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Dec

embe

r 5,

202

1