�2-HS Glycoprotein/Fetuin, a Transforming Growth Factor-�/BoneMorphogenetic Protein Antagonist, Regulates Postnatal BoneGrowth and Remodeling*

Received for publication, December 20, 2001, and in revised form, March 18, 2002Published, JBC Papers in Press, March 18, 2002, DOI 10.1074/jbc.M112234200

Melanie Szweras‡§¶, Danmei Liu‡, Emily A. Partridge‡, Judy Pawling‡, Balram Sukhu‡,Cameron Clokie�, Willi Jahnen-Dechent**, Howard C. Tenenbaum‡�, Carol J. Swallow‡,Marc D. Grynpas‡ ‡‡, and James W. Dennis‡§ §§

From the ‡Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, Ontario M5G 1X5,the §Department of Molecular and Medical Genetics, the �Department of Dentistry, and the ‡‡Departmentof Laboratory Medicine and Pathology, University of Toronto, Toronto, Ontario M5S 1A8, Canada,and **IZKF BIOMAT, Klinikum der RWTH Aachen, Pauwelsstrasse 30, 52057 Aachen, Germany

Soluble transforming growth factor-� (TGF-�)/bonemorphogenetic protein (BMP)-binding proteins arewidely distributed in mammalian tissues and controlcytokine access to membrane signaling receptors. Theserum and bone-resident glycoprotein �2-HS-glycopro-tein/fetuin (ASHG) binds to TGF-�/BMP cytokines andblocks TGF-�1 binding to cell surface receptors. There-fore, we examined bone growth and remodeling pheno-types in ASHG-deficient mice. The skeletal structure ofAhsg�/� mice appeared normal at birth, but abnormali-ties were observed in adult Ahsg�/� mice. Maturation ofgrowth plate chondrocytes was impaired, and femurslengthened more slowly between 3 and 18 months of agein Ahsg�/� mice. However, bone formation was in-creased in Ahsg�/� mice as indicated by greater corticalthickness, accelerated trabecular bone remodeling, andincreased osteoblast numbers on bone surfaces. Thenormal age-related increase in cortical thickness andbone mineral density was accelerated in Ahsg�/� miceand was associated with increased energy required tofracture. Bone formation in response to implanted BMPcytokine extended further from the implant in Ahsg�/�

compared with Ahsg�/� mice, confirming the interactionbetween ASHG and TGF-�/BMP cytokines in vivo. Ourresults demonstrate that ASHG blocks TGF-�-depend-ent signaling in osteoblastic cells, and mice lackingASHG display growth plate defects, increased bone for-mation with age, and enhanced cytokine-dependentosteogenesis.

ASHG1 or fetuin is a liver secretory glycoprotein found athigh levels in serum and mineralized bone (1, 2). The humanAHSG gene is located on chromosome 3q27, and two commonallelic forms have been identified that correlate with differentbone phenotypes (3). Homozygosity for AHSG*1 allele is asso-

ciated with shorter stature and reduced bone quality measuredby calcaneal broadband ultrasound (4). Further interest stemsfrom the observations that elevated serum AHSG correlateswith a common form of mild osteogenesis imperfecta (5),whereas depressed levels are observed in Paget’s disease, acondition characterized by increased bone turnover (6). AHSGis also a reverse acute-phase reactant, as serum levels declineby 20–30% during acute inflammation (7).

ASHG protein has two cystatin-like domains with a char-acteristic arrangement of disulfide loops (8), a calcium phos-phate-binding site near the N terminus (9), and a TGF-�cytokine-binding motif (10). The latter is a disulfide-loopedsequence in the N-terminal cystatin domain from Cys-96 toCys-114 (human sequence) which shares homology with theextracellular domain of TGF-� receptor type II (T�RII)(Cys-84 to Cys-101). In surface plasmon resonance assays,these peptides bind to TGF-� and BMP cytokines with spec-ificity characteristic of native ASHG and T�RII, suggestingthey are the major cytokine-binding domains in each glyco-protein (10). ASHG binds to TGF-� cytokines with KD valuesranging from 0.03 to 2.0 �M for BMP-2, BMP-4, BMP-6,TGF-�1, and TGF-�2 in rank order. Furthermore, ASHGblocked TGF-�1 and BMP-2 anti-proliferative and osteogenicactivities in cell culture (10). Although the affinity of theseinteractions is modest, the serum concentration of ASHG is�12 �M, and based on mass action, ASHG would be expectedto influence cytokine availability in vivo. The on and off ratesfor ASHG-cytokine binding are slow (10), a property charac-teristic of other cytokine antagonists (11). EndogenousTGF-�1 in rat bone marrow cell cultures is required fordifferentiation, but at �10 pM TGF-�1 completely inhibitsmineralization. Osteogenesis was observed to be maximal at�1 pM TGF-�1, and as might be expected, addition of eitherASHG or neutralizing anti-TGF-�1 antibodies increased therequired TGF-�1. Therefore, osteogenesis depends on a molarratio of ASHG to cytokine, suggesting that ASHG may estab-lish morphogenic fields for osteo-induction and thereby con-trol bone remodeling. Osteogenesis in rat bone marrow cellcultures was optimal at �1/300,000 molar ratio of TGF-�1 toASHG which reflects the relative physiological levels andactivities of these proteins (12).

Spatially distinct sources of cytokine and antagonist areknown to establish gradients of cytokine activity and therebycontrol regional differentiation in the early embryo (13, 14). Forexample, noggin and chordin establish a BMP-4 gradient in thegastrula stage vertebrate embryo (15). Noggin is also expressed

* This work was supported in part by grants from Canadian Insti-tutes for Health Research (to J. W. D. and M. G.) and the PhysiciansServices Inc. (to C. J. S.). The costs of publication of this article weredefrayed in part by the payment of page charges. This article musttherefore be hereby marked “advertisement” in accordance with 18U.S.C. Section 1734 solely to indicate this fact.

¶ Supported by a Canadian Institutes for Health Researchstudentship.

§§ To whom correspondence should be addressed. Tel.: 416-586-8233,Fax: 416-586-8588; E-mail: Dennis@mshri.on.ca.

1 The abbreviations used are: ASHG, �2-HS-glycoprotein; TGF-�,transforming growth factor-�; BMP, bone morphogenetic protein;T�RII, TGF-� receptor, type II; BMD, bone mineral density.

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 277, No. 22, Issue of May 31, pp. 19991–19997, 2002© 2002 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

This paper is available on line at http://www.jbc.org 19991

in condensing cartilage, and noggin-deficient embryos displayabnormal growth plates and joints (16). Other TGF-� cytokine-binding proteins include serum �2-macroglobulin (17), solublebetaglycan (18), matrix proteoglycans, decorin, biglycan, andfibromodulin (19–21). Biglycan-deficient mice display an age-dependent low bone mass (22), demonstrating that matrix-localized cytokine-binding proteins can influence bonehomeostasis.

Skeletogenesis precedes expression and accumulation ofASHG in mineralizing bone, and indeed Ahsg�/� mice displayno obvious anatomical abnormalities at birth (23). However,bone is continuously remodeled during adult life, a processwhereby osteoclasts degrade the bone and release cytokines, inturn stimulating osteoblast recruitment from bone marrow torestore the bone. It is possible that bone and serum-derivedASHG bind the released cytokines near the remodeling surfaceand regulate osteogenic activity. Here we have examined femurhistology and geometry, as well as bone structural and dynamicproperties in young and old Ahsg�/� mice and their ASHG-expressing littermates. Bone formation was increased andmaximum load enhanced in Ahsg�/� mice, whereas chondro-genesis and longitudinal bone growth were impaired. Further-more, ectopic bone formation in response to intramuscularBMP cytokine implants was enhanced in ASHG-deficient mice,consistent with loss of a negative regulator of osteogenesis.TGF-�/BMP cytokines are known to accumulate in mineralizedbone and contribute to chondrogenesis and osteogenesis (24–27). Furthermore, an age-related decline in bone TGF-�1 cor-relates with loss of bone quality (28, 29) Our results demon-strate that ASHG is a cytokine antagonist and regulatesgrowth plate chondrogenesis and osteogenesis in remodelingbone. These results also suggest that an age-related imbalancebetween cytokine and antagonist levels may lead to bonediseases.

EXPERIMENTAL PROCEDURES

ASHG Mutant Mice—ASHG-deficient mice were generated by tar-geted gene mutation in embryonic cells, which removed the entire

coding region of the gene as described previously (23). An equal distri-bution of male and female mice with Ahsg�/�, Ahsg�/�, and Ahsg�/�

genotypes on a 129/sv � C57BL/6 background were generated for theanalysis, and the data from both sexes were averaged.

Immunohistochemistry—Dissected femurs were fixed in 10% forma-lin for 7 days, decalcified in 10% formic acid, then dehydrated througha series of ethanol (20–100%) and xylene solutions, and finally embed-ded in paraffin (Paraplast X-Tra). Blocks were softened with mollifex(BDH) and then sectioned. The sections were deparaffinized, hydratedthrough xylene and ethanol, incubated with a 1/500 dilution of rabbitanti-mouse ASHG antibody in 1.5% serum in phosphate-buffered sa-line, washed, then incubated with 1/5000 dilution of biotinylated anti-rabbit antibody, and developed with the horseradish peroxidase sub-strate 3,3�-diaminobenzidine tetrahydrochloride (Vector Laboratories).The slides were counterstained with hematoxylin.

Bone Histomorphometry and Dynamic Properties—Three-month-oldmice were injected in the tail vein with 10 mg of tetracycline/kg at 10and 3 days prior to sacrifice. The femurs were dissected and cut coro-nally, and both halves were fixed in 70% ethanol, dehydrated throughincreasing strengths of acetone, infiltrated in increasing strengths ofspurr/acetone solutions, and embedded undecalcified in polymerizedplastic spurr blocks. Blocks were cut into a series of three 5-�m sectionsand one 7-�m section. The 7-�m section was left unstained for dynamicmorphometry. Unstained sections were examined with a fluorescentsource and the Bioquant OS2 software to measure length of single label,length of double label, and distance between labels.

The 5-�m section next in series to the unstained section was stainedusing a modified Masson-Goldner trichrome technique that permitsoptimal discrimination between mineralized and non-mineralized (os-teoid) bone (30). The next section in the series was stained with tolui-dine blue to enumerate osteoblasts on trabecular bone surfaces andadipocytes in the marrow. The final section in the series was stainedfor tartrate-resistant acid phosphatase to enumerate osteoclasts. Sec-tions were viewed with a semi-automatic image analyzer (Leitz ASM,Wetzlar, France) and quantified using an IBM-PC microcomputerand Bioquant image analysis software.

Static and dynamic parameters of trabecular bone were analyzed byhistomorphometric methods, which complied with the nomenclature,and were calculated according to the ASBMR Histomorphometric No-menclature Committee (31). Data for male and female mice calculatedseparately showed the same relative differences.

Mechanical Testing and Dual Energy X-ray Absorptiometry—De-structive three-point bending was performed on the right femurs of

FIG. 1. ASHG is localized to mineralized bone. Sections of femur from Ahsg�/� (A), Ahsg�/� (B), and Ahsg�/� (C) mice stained withantibodies to ASHG. D–F, higher power images from Ahsg�/� mice revealing ASHG protein in bone immediately surrounding growth plate (D),in trabecular bone (E), and in mineralized bone adjacent to osteocytes (F). Abbreviations used are as follows: ac, articular cartilage; gp, growthplate; c, cortical bone; m, marrow; t, trabeculae; and o, osteocytes.

�2-HS Glycoprotein and TGF-�-dependent Osteogenesis19992

mice using a screw-driven mechanical testing machine (Instron model1011, Canton, MA). Each bone was placed on two supports spaced 6.7mm apart, and a load was applied to the bone midway between thesupports at a deformation rate of 1 mm/min. From the load displace-ment curve, the maximum load and maximum displacement weremeasured, and the stiffness was determined from a linear regression ofthe initial portion of the curve. The length, diameter, and corticalthickness of the bones were determined using digital calipers. DualEnergy x-ray absorptiometry (Pixi Mus, Lunar Corp., Madison, WI) wasused to measure bone mineral content, bone area, and bone mineraldensity (BMD) of femurs.

Osteogenesis Induction—Femoral bones were removed under asepticconditions from adult male Wistar rats (120 g), cleaned of adherent softtissues, and washed extensively in antibiotics. The distal ends wereremoved, and the marrow contents were flushed out with 10 ml ofculture medium. The cells were dispersed by repeated passage througha 20-gauge needle and incubated in �-minimum Eagle’s medium sup-plemented with 15% fetal bovine serum, ascorbic acid (50 �g/ml), anti-biotics (100 �g/ml penicillin G, 50 �g/ml gentamicin, 0.3 �g/ml fungi-zone), 10 mM �-glycerophosphate, 10�8 M dexamethasone, and vitaminC. Following 6 days of culture, the cells were re-plated at a density of1 � 102 cells/mm2 in 96-well plates, and grown for another 12–14 dayswithout dexamethasone and with 10 nm recombinant BMP-2. At theend of culture, the cells were fixed with 10% buffered formalin andstained for calcium with Alizarin Red-s to identify mineralized bonenodules. To quantify mineralized tissue formation in the cultures, theabsorbance at 525 nm was measured using a 96-well plate reader.

For the analysis of ectopic bone formation, cytokine embedded ingelatin capsules was implanted into the thigh muscle of mice, using 12mice per genotype. The capsules contained 5 mg of native human BMPcytokine, extracted and purified from demineralized human bone ma-trix as described previously (32).

RESULTS

Growth Plate Defects and Reduced Femur Length in Ahsg�/�

Mice—Serum ASHG was reduced by �50% in Ahsg�/� andabsent in the Ahsg�/� mice as reported previously (23). ASHGprotein was concentrated in the mineralized regions of bones(2) from Ahsg�/� and Ahsg�/� mice (Fig. 1, A and B) and wasparticularly concentrated in bone surrounding the growth plate(Fig. 1D). The density of ASHG protein displayed heterogeneityin trabecular bone, possibly due to variations in when the bonewas last remodeled (Fig. 1E). ASHG protein was more concen-trated in bone surrounding osteocytes (Fig. 1F), cells that areencased in bone and respond to mechanical stress by secretingparacrine factors that stimulate bone remodeling (33).

FIG. 2. Epiphyseal growth plate morphology is disrupted inAhsg�/� mice. Sections of femurs from Ahsg�/� (A) and Ahsg�/� (B)mice were stained with toluidine blue to examine growth plate integ-rity. C, growth plate discontinuities. D, cartilage islands. Sections fromAhsg�/� (E) and Ahsg�/� (F) mice were stained with trichrome toreveal chondrocyte organization. G, columns of chondrocytes per growthplate; H, hypertrophic chondrocytes per total cell number in the growthplate. The data represents the mean � S.E. of 12 mice per genotype, and*, p � 0.01 versus Ahsg�/�. Abbreviations used are as follows: gp,growth plate; ci, cartilage islands; cc, chondrocyte columns; and hc,hypertrophic chondrocyte.

FIG. 3. Geometric and structural properties of femurs inAhsg�/�, Ahsg�/�, and Ahsg�/� ice. The length (A), cortical thickness(B), bone mineral content (C), bone mineral density (D), maximum load(failure) (E), and energy (F) were measured in Ahsg�/�, Ahsg�/�, andAhsg�/� mice at 3–4 and 12–18 months. The energy and load weredetermined by three-point bending of femurs. Results are the mean �S.E. of 8–10 mice per genotype, and *, p � 0.05 versus Ahsg�/� mice.

�2-HS Glycoprotein and TGF-�-dependent Osteogenesis 19993

Because polymorphic differences in the coding region of hu-man AHSG correlate with differences in stature, we examinedfemur growth plate histology and dimensions in the mice. Bal-anced groups of littermates representing all three genotypeswere used throughout our studies. The femur growth plates in3-month-old Ahsg�/� mice were fragmented, and chondrocyteswithin the growth plate were poorly organized (Fig. 2, A and B).Twice as many discontinuities or breaks in growth plates wereobserved in Ahsg�/� compared with Ahsg�/� mice, and carti-lage islands in the metaphysis were 6-fold more frequent inAhsg�/� mice (Fig. 2, C and D). The cartilage islands failed tocalcify as indicated by lack of von Kossa staining, which couldindicate a failure of osteoclast remodeling (data not shown).Chondrocytes are normally arranged in vertical columns in thegrowth plate, where they undergo a spatially precise programof differentiation regulated by Indian hedgehog (Ihh), parathy-roid hormone-related peptide, and BMP cytokines (34–36).Chondrocyte appeared disorganized (Fig. 2, E and F), and thenumber of columns per growth plate was reduced by 53% inAhsg�/� and unchanged in Ahsg�/� compared with wild typemice (Fig. 2G). Hypertrophic chondrocytes per growth platewere reduced in Ahsg�/� mice but significantly increased inAhsg�/� mice (Fig. 3H), whereas total chondrocyte cell countsper growth plate were not significantly different between gen-otypes (data not shown).

Endochondral ossification by maturing chondrocytes occursat the proximal aspect of the epiphyseal growth plate anddrives longitudinal bone growth. Accordingly, the deficiency ingrowth plate chondrocyte maturation was associated with re-duced longitudinal bone growth in Ahsg�/� mice. Femurlength in Ahsg�/� mice compared with wild type age-matchedmice was reduced by 9% at 3–4 months and by 14.7% at 12–18months (Fig. 3A). However, femoral cortical thickness was sig-nificantly increased in the Ahsg�/� mice at both 3–4 and 12–18months of age, indicating an increase in osteoblastic activityrelative to osteoclastic activity in Ahsg�/� mice (Fig. 3B). Het-erozygous mice were not significantly different from wild typefor these measurements.

BMD and Strength Increase with Age in Ahsg�/� Mice—Changes in cortical thickness are expected to alter the mechan-ical and structural properties of bone. Bone mineral contentand bone area were determined using dual energy x-ray ab-sorptiometry analysis of mice at 3–4 months and at 12–18months of age. The Ahsg�/� mice displayed no significantchange in bone mineral content, but BMD, which is bone min-eral content normalized to bone area, was increased by 15% inthe older Ahsg�/� mice as compared with wild type and het-erozygous littermates (Fig. 3, C and D). Mechanical propertiesof femurs were measured at 3–4 and 12–18 months of age usinga three-point bending test. Maximum load normally increaseswith age as we observed here for all genotypes. No differencesbetween genotypes were observed in young mice, but the olderAhsg�/� mice displayed a higher maximum load, consistentwith greater BMD and femoral cortical thickness in these mice(Fig. 3E). The energy to fracture normally decreases with age,but this value increased significantly in the older Ahsg�/� mice(Fig. 3F).

Osteogenesis and Adipogenesis Is Enhanced in Bone Marrowof Ahsg�/� Mice—TGF-�/BMP can induce progenitor cells inbone marrow cultures to differentiate along different cell lin-eages depending on the cytokine concentrations (12, 25). Adi-pocytes in the bone marrow of Ahsg�/� mice were increased5-fold compared with wild type and heterozygous mice (Fig. 4,A and B). Osteoblasts per trabecular surface were also in-creased by 60% in Ahsg�/� mice, whereas the osteoclasts pertrabecular surface were not significantly different (Fig. 4, Cand D). These observations suggest that progenitor cell recruit-ment along adipogenic and osteogenic lineages occurs at ahigher frequency in Ahsg�/� mice compared with age-matchedAhsg�/� and Ahsg�/� mice. Furthermore, increased osteoblastcontent and cortical bone thickness indicate that net boneformation rates were enhanced in Ahsg�/� mice (Fig. 3B andFig. 4C).

To measure directly the dynamic parameters of bone remod-eling, newly forming bone was pulse-labeled with two intrave-nous injections of tetracycline administered 7 days apart (Fig.4E). The mineralizing surface and mineral formation rateswere increased 50 and 100% in Ahsg�/� mice compared withAhsg�/� and Ahsg�/� mice, respectively (Fig. 4, F and G),consistent with observed differences in osteoblasts/surface(Fig. 4C). Surprisingly, mineralizing surface and mineral for-mation rates were significantly decreased in Ahsg�/� micecompared with control mice (Fig. 4, F and G). Correspondingly,mineralization lag time was decreased in Ahsg�/� and in-creased in Ahsg�/� compared with wild type mice (Fig. 4H).

The histomorphometric measurements of trabecular bone at3 months of age also indicate an unusual relationship betweenthe ASHG genotypes. Although Ahsg�/� and Ahsg�/� micewere similar, the heterozygous mice differed significantly forseveral parameters. Trabecular number and trabecular surface

FIG. 4. Osteogenesis, adipogenesis, and bone remodeling isaltered in Ahsg�/�, Ahsg�/�, and Ahsg�/� mice. A, toluidine bluestaining of femoral sections reveals adipocytes in marrow as whiteareas. B, adipocytes per field, osteoblasts per bone surface, and oste-oclasts per bone surface were quantified. E, dynamic properties of boneremodeling in Ahsg�/�, Ahsg�/�, and Ahsg�/� mice were examined bydual tetracycline labeling. The arrows mark the two labeled frontsobserved by fluorescence microscopy. F–H, mineralizing surface/bonesurface, mineral formation rate, and mineralization lag time weremeasured. Results are the mean � S.E. of 12 Ahsg�/�, 8 Ahsg�/�, and12 Ahsg�/� mice, and *, p � 0.005 versus Ahsg�/�.

�2-HS Glycoprotein and TGF-�-dependent Osteogenesis19994

were increased and trabecular separation reduced in Ahsg�/�,whereas total trabecular bone volume was not significantlydifferent between genotypes (Table I). Therefore, the trabecu-lar lattice in Ahsg�/� mice appeared to be of a finer meshworkthan that observed in either wild type or ASHG-deficient mice.

Enhanced Ectopic Bone Formation in Ahsg�/� Mice—ASHGbinds to BMP-2 with a KD of 10�8 M (10) and inhibited BMP-2-stimulated osteogenesis in rat bone marrow cell cultures (Fig.5A). Lacking the antagonist activity, ASHG-deficient micewere expected to be more susceptible to the bone morphogenicactivity of exogenous cytokine. To test the hypothesis in vivo,

mice were implanted intramuscularly with pellets containingbone morphogenic cytokines, and ectopic bone formation wasmeasured 4 weeks later. Mineral content and mineralized areaat the implant site were significantly greater in Ahsg�/� andAhsg�/� mice compared with Ahsg�/� mice (Fig. 5, B and C).The mineralized area displayed the morphology of bone, withosteoblasts on the bone surface, encased osteocytes, and mar-row-like compartments (Fig. 5, D and E). Thus, exogenouscytokine induced a greater area of ectopic osteogenesis inAhsg�/� and Ahsg�/� mice, confirming an interaction betweenASHG and cytokine in the regulation of osteogenesis in vivo.

FIG. 5. ASHG antagonizes BMP-stimulated osteogenesis in vitro and in vivo. A, bone marrow cells were cultured in the presence of 10nM recombinant BMP-2 for 14 days, and mineralization in the cultures was detected by Alizarin Red-s staining. B–E, littermates at 3 month of agereceived intramuscular implants of native BMP cytokine (5 mg), and bone formation was monitored 4 weeks later. B, mineral content. C,mineralized area was measured by dual energy x-ray absorptiometry. The results are the mean of 10 mice per group � S.E. *, p � 0.05 versusAhsg�/�. D, sections of mineralized tissues in Ahsg�/� mice at the site of BMP cytokine implants, displaying morphology characteristic oftrabecular bone marrow. E, higher power image of ectopic bone. Abbreviations used are as follows: m, muscle; b, bone. The open arrow marks anosteocyte encased in bone, and the solid arrow marks osteoblasts on bone surface.

TABLE IStatic parameters of femoral trabecular bone from Ahsg�/�, Ahsg�/�, and Ahsg�/� mice at 3 months

Femurs were stained with trichrome, and histomorphometric measurements were taken on eight fields per section from 12 Ahsg�/�, 8 Ahsg�/�,and 12 Ahsg�/� mice. Data are mean � S.E.

Parameter Ahsg�/� Ahsg�/� Ahsg�/�

Trabecular bone volume (%) 11.25 � 0.94 13.03 � 1.04 12.96 � 1.40Mineralized trabecular bone volume (%) 10.73 � 0.88 12.58 � 1.01 12.38 � 1.42Osteoid volume (%) 4.57 � 0.55 3.48 � 0.81 5.06 � 1.01Osteoid surface (%) 25.60 � 2.98 17.50 � 2.59 28.10 � 2.84a

Osteoid thickness (�m) 3.02 � 0.22 3.08 � 0.47 3.25 � 0.23Trabecular thickness (�m) 34.73 � 2.12 31.77 � 1.22 40.06 � 2.47a

Eroded surface (%) 0.70 � 0.12 0.58 � 0.13 0.59 � 0.09Trabecular number (mm�1) 3.22 � 0.17b 4.04 � 0.25 3.20 � 0.19a

Trabecular separation (�m) 287.04 � 20.25b 223.02 � 22.84 283.56 � 17.70a

Total trabecular surface (�m) 6735.19 � 413.71b 8407.51 � 558.40 6392.47 � 447.50a

Total osteoid surface (�m) 1645.10 � 196.45 1430.77 � 276.55 1769.71 � 193.18a p � 0.05 between Ahsg�/� and Ahsg�/�.b p � 0.05 between Ahsg�/� and Ahsg�/� groups, by Student’s t test.

�2-HS Glycoprotein and TGF-�-dependent Osteogenesis 19995

DISCUSSION

ASHG is a TGF-�/BMP antagonist that localizes to mineral-ized bone, and here we have examined ASHG-deficient mice forbone defects that might be associated with a hyperactive cyto-kine environment. Chondrocyte differentiation and organiza-tion in the growth plate was impaired, and this was accompa-nied by slower longitudinal bone growth in Ahsg�/� micecompared with Ahsg�/� and Ahsg�/� mice. An excess of cyto-kine activity in the Ahsg�/� growth plate may cause thisphenotype, as exogenous TGF-�1 has been shown to preventterminal differentiation of cells in growth plate explants intohypertrophic chondrocytes (37). The disruption of chondrocyteorganization in the epiphyseal growth plate of Ahsg�/� micemay also be due in part to a defect in positioning of prechon-drocytes into columnar structures. In this regard, mice lackingthe BMP antagonist noggin are defective in joint formation dueto an apparent failure of articular chondrocytes to positioncorrectly (16). As cells move toward a source of agonist, thegradient strength can improve the accuracy of cell chemotaxisand achieve precise positioning of the cell at the destination.An antagonist primarily limits the effective range of an ago-nist, but when the two emanate from different positions, theantagonist effectively increases the gradient strength of theagonist (38). Importantly, ASHG-deficient mice were more sus-ceptible to exogenous bone morphogenic cytokines, consistentwith the absence of a BMP/TGF-� antagonist in vivo. The areaof ectopic osteogenesis was larger in both Ahsg�/� andAhsg�/� mice, indicating that BMP activity extended furtherfrom the implant site when ASHG was depleted.

BMP-2 can induce differentiation of mesenchymal cells intoboth osteoblasts and adipocytes in proportions that depend uponcytokine receptor expression and growth conditions (39). Bothosteoblasts on trabecular bone and adipocytes in bone marrowwere markedly increased in Ahsg�/� mice, indicating an en-hanced recruitment of bone marrow mesenchymal cell precur-sors. This was associated with increased trabecular bone remod-eling in Ahsg�/� mice, a progressive increase in cortical bonethickness, BMD, and greater maximum load to fracture in oldermice. In an earlier study, Ahsg�/� mice were found to be moresusceptible to spontaneous soft tissue mineralization (23) andattributed to the direct inhibition of hydroxyapatite formation byASHG. However, ASHG blocks differentiation well before min-eralization in bone marrow cultures, as indicated by a lack ofosteocalcin, osteopontin, and alkaline phosphatase gene expres-sion (12). Furthermore, addition of ASHG to the bone marrowcultures after differentiation, but prior to mineralization, did notinhibit the latter. The ectopic bone formed in response to bonemorphogenic cytokines in Ahsg�/� mice displayed the typicalbone morphology, consistent with an osteogenic process ratherthan accumulation of soft tissue calcification. However, it is pos-sible that ASHG regulates both cytokine-dependent osteogenesisand the final stage of mineralization.

The bone phenotype in Ahsg�/� mice has features that arecomparable with genetic mutations affecting TGF-� cytokines.Missense mutations of TGF-�1 latency-associated peptidecause enhanced activation of TGF-�1 in Camurati-Engelmanndisease, an autosomal dominant disorder characterized by hy-perosteosis and sclerosis of the diaphysis of the long bones (40).BMP-3 has recently been shown to antagonize the osteogenicBMPs, and BMP3�/� mice show a 2-fold increased in trabec-ular bone (41). TGF-�1-deficient mice display reduced bonemass and elasticity, as well as growth plate defects (42). Fur-thermore, a polymorphism in the coding region of the humanTGF-�1 gene has been correlated with decreased serum levelsof TGF-�1 and susceptibility to osteoporosis in postmenopausalJapanese women (43). Interestingly, transgenic mice express-

ing TGF-�2 in osteoblasts show bone loss (44), and mice ex-pressing dominant negative T�RII in osteoblasts display anincrease in trabecular bone (45). However, the phenotypes arecomplex, as TGF-�2 transgenic mice also show increased osteo-blast and osteocyte differentiation, which may be in part de-pendent on TGF-�-mediated increases in osteoclastic activity(46).

TGF-�1 is a negative regulator of lymphocyte proliferationand inflammation (47), and therefore loss of systemic antago-nists might be expected to result in immune suppression. Micelacking �2-macroglobulin, another serum TGF-�-binding glyco-protein, display a hypo-inflammatory phenotype and resistanceto endotoxin challenge (48). Ahsg�/� mice are also more resist-ant to endotoxin than Ahsg�/� mice.2 T cells from Ahsg�/� miceare less responsive to stimulation by anti-CD3 and anti-CD28antibodies, and in addition, skin inflammation induced by top-ical application of arachadonic acid was reduced and returnedto normal more quickly in Ahsg�/� compared with Ahsg�/�

mice. Interestingly, Ahsg gene expression is down-regulated byinterleukin-1 and interleukin-6 in hepatic cells (49), and ASHGis known to be a reverse acute-phase reactant (7). It is possiblethat down-regulation of ASHG enhances TGF-�-mediated im-mune suppression. Elevated interleukin-6 is also associatedwith osteoporosis in inflammatory bowel disease due to in-creased osteoclastic activity (50); an associated reduction inbone ASHG could also affect osteogenesis.

For some femur characteristics, ASHG heterozygous micewere not intermediate between Ahsg�/� and Ahsg�/� mice,notably maturation of growth plate chondrocytes, static param-eters of trabecular bone, and dynamic measures of bone remod-eling. One interpretation is that ASHG can exert both a posi-tive and a negative effect that influence these phenotypes. Forexample, a cytokine buffering or carrier activity might be re-flected in the observed phenotype of Ahsg�/� mice, whereasloss of the cytokine antagonist activity dominates the observedphenotype in the Ahsg�/� bone. In spatial terms, the concen-tration of ASHG and cytokines in bone and their release withremodeling could establish cytokine gradients in the bone mar-row. The Ahsg�/� mice would lack the ASHG-dependent cyto-kine gradient, resulting in an enlarged zone of stromal cellrecruitment, which is consistent with the observed increase inosteoblast and adipocyte content throughout the marrow ofAhsg�/� mice (Fig. 4). A 2–3-fold reduction in serum ASHGand near wild type levels of ASHG in mineralized bone ofAhsg�/� mice should result in a steeper ASHG gradient and,consequently, a more narrow region where cytokine levels areoptimal for osteogenesis. Consistent with this model, trabecu-lar bone in Ahsg�/� mice displayed a significantly finer mesh-work as well as reduced remodeling rates compared with thatin either Ahsg�/� or Ahsg�/� mice (Table I). However, theinteractions are likely to be more complex, as ASHG binds tomultiple TGF-�/BMP cytokines with different affinities andtherefore has the potential to change cytokine activities in acomplex manner that could readily give rise to a distinctAhsg�/� phenotype. In any event, detection of an Ahsg�/�

phenotype suggests that a relatively modest change in ASHGlevels can affect bone remodeling rates and trabeculararchitecture.

Bone TGF-� levels decline with age (28, 29), which may leadto an imbalance with antagonists and loss of cytokine-depend-ent stromal cell recruitment. However, TGF-�1 cytokine ex-pression in tissue fibroblasts increases with age (51), and animbalance relative to antagonists may promote ectopic osteo-

2 E. A. Partridge, M. Szweras, C. J. Swallow, and J. W. Dennis et al.,manuscript in preparation.

�2-HS Glycoprotein and TGF-�-dependent Osteogenesis19996

genesis, immune suppression, and fibrosis in the elderly.TGF-�1 contributes to pathologies associated with atheroscle-rosis, kidney disease, chronic obstructive pulmonary disease inasthma patients, diabetic nephropathy, vitreoretinopathy, scarformation in wound healing (52, 53), and also tumor progres-sion (54). We have observed that intestinal tumor progressionwas enhanced in Ahsg�/� compared with Ahsg�/� mice with anApc mutation (Min/�).3 Thus, TGF-� antagonists or blockers ofcytokine signaling may have therapeutic applications in ad-vanced cancers and fibrosis. Lending further support to thisnotion, genetic deletion of Smad3, which mediates TGF-� sig-nal transduction in lymphoid and stromal cells, speeds woundhealing, reduces scarring, and prevents radiation-induced fi-brosis in mice (55). Based on the cytokine antagonist activity ofASHG and the Ahsg�/� phenotype, recombinant ASHG mightbe useful in the treatment of diseases where TGF-� overexpres-sion adversely affects outcome.

Acknowledgments—We thank M. Kasra, M. Mendes, M. Cui, B.Rittenberg, and S. Barkin for technical assistance and C. E. Warren forhelpful discussion.

REFERENCES

1. Pedersen, K. O. (1944) Nature 3914, 575–5752. Dickson, I. R., Poole, A. R., and Veis, A. (1975) Nature 256, 430–4323. Sebetan, I. M., and Heshmat, M. M. (1988) Z. Rechtsmed. 101, 205–2074. Zmuda, J. M., Eichner, J. E., Ferrell, R. E., Bauer, D. C., Kuller, L. H., and

Cauley, J. A. (1998) Calcif. Tissue Int. 63, 5–85. Dickson, I. R., Bagga, M., and Paterson, C. R. (1983) Calcif. Tissue Int. 35,

16–206. Ashton, B. A., and Smith, R. (1980) Clin. Sci. (Lond.) 58, 435–4387. Lebreton, J. P., Joisel, F., Raoult, J. P., Lannuzel, B., Rogez, J. P., and

Humbert, G. (1979) J. Clin. Invest. 64, 118–1298. Kellermann, J., Haupt, H., Auerswald, E.-A., and Muller-Esterl, W. (1989)

J. Biol. Chem. 264, 14121–141289. Schinke, T., Amendt, C., Trindl, A., Poschke, O., Muller-Esterl, W., and

Jahnen-Dechent, W. (1996) J. Biol. Chem. 271, 20789–2079610. Demetriou, M., Binkert, C., Sukhu, B., Tenenbaum, H. C., and Dennis, J. W.

(1996) J. Biol. Chem. 271, 12755–1276111. Schneyer, A. L., Rzucidlo, D. A., Sluss, P. M., and Crowley, W. F., Jr. (1994)

Endocrinology 135, 667–67412. Binkert, C., Demetriou, M., Sukhu, B., Szweras, M., Tenenbaum, H. C., and

Dennis, J. W. (1999) J. Biol. Chem. 274, 28514–2852013. Neumann, C., and Cohen, S. (1997) Bioessays 19, 721–72914. Hogan, B. L. M. (1996) Genes Dev. 10, 1580–159415. Jones, C. M., and Smith, J. C. (1998) Dev. Biol. 194, 12–1716. Brunet, L. J., McMahon, J. A., McMahon, A. P., and Harland, R. M. (1998)

Science 280, 1455–145717. O’Connor-McCourt, M. D., and Wakefield, L. M. (1987) J. Biol. Chem. 262,

14090–1409918. Lopez-Casillas, F., Payne, H. M., Andres, J. L., and Massague, J. (1994) J. Cell

Biol. 124, 557–56819. Yamaguchi, Y., Mann, D. M., and Ruoslahti, E. (1990) Nature 346, 281–28420. Hildebrand, A., Romaris, M., Rasmussen, L. M., Heinegard, D., Twardzik,

D. R., Border, W. A., and Ruoslahti, E. (1994) Biochem. J. 302, 527–53421. Piek, E., Heldin, C. H., and ten Dijke, P. (1999) FASEB J. 13, 2105–212422. Xu, T., Bianco, P., Fisher, L. W., Longenecker, G., Smith, E., Goldstein, S.,

Bonadio, J., Boskey, A., Heegaard, A.-M., Sommer, B., Satomura, K.,Dominguez, P., Zhao, C., Kulkarni, A. B., Gehron Robey, P., and Young,M. F. (1998) Nat. Genet. 20, 78–82

23. Jahnen-Dechent, W., Schinke, T., Tridl, A., Muller-Esterl, W., Sablizky, F.,Kaiser, S., and Blessing, M. (1997) J. Biol. Chem. 272, 31496–31503

24. Pelton, R. W., Saxena, B., Jones, M., Moses, H. L., and Gold, L. I. (1991) J. CellBiol. 115, 1091–1105

25. Joyce, M. E., Roberts, A. B., Sporn, M. B., and Bolander, M. E. (1990) J. CellBiol. 110, 2195–2207

26. Pinero, G. J., Farach-Carson, M. C., Devoll, R. E., Aubin, J. E., Brunn, J. C.,and Butler, W. T. (1995) Arch. Oral Biol. 40, 145–155

27. Centrella, M., McCarthy, T. L., and Canalis, E. (1991) J. Bone Jt. Surg. Am.73, 1418–1428

28. Pfeilschifter, J., Diel, I., Scheppach, B., Bretz, A., Krempien, R., Erdmann, J.,Schmid, G., Reske, N., Bismar, H., Seck, T., Krempien, B., and Ziegler, R.(1998) J. Bone Miner. Res. 13, 716–730

29. Gazit, D., Zilberman, Y., Ebner, R., and Kahn, A. (1998) J. Cell. Biochem. 70,478–488

30. Villanueva, A. (1988) J. Histotechnol. 11, 249–25131. Parfitt, A. M., Drezner, M., Glorieux, F., Kanis, J. A., Malluche, H., Meunier,

P. J., Ott, S. M., and Recker, R. R. (2000) J. Bone Miner. Res. 2, 597–61032. Mizutani, H., and Urist, M. R. (1982) Clin. Orthop. 171, 213–22333. Huiskes, R., Ruimerman, R., van Lenthe, G. H., and Janssen, J. D. (2000)

Nature 405, 704–70634. Bailon-Plaza, A., Lee, A. O., Veson, E. C., Farnum, C. E., and van der Meulen,

M. C. H. (1999) Bone 24, 211–21635. St-Jacques, B., Hammerschmidt, M., and McMahon, A. P. (1999) Genes Dev.

13, 2072–208636. Karaplis, A. C., Luz, A., Glowacki, J., Bronson, R. T., Tybulewicz, V. L. J.,

Kronenberg, H. M., and Mulligan, R. C. (1994) Genes Dev. 8, 277–28937. Ballock, R. T., Heydemann, A., Wakefield, L. M., Flanders, K. C., Roberts,

A. B., and Sporn, M. B. (1993) Dev. Biol. 158, 414–42938. Barkai, N., Rose, M. D., and Wingreen, N. S. (1998) Nature 396, 422–42339. Ji, X., Chen, D., Xu, C., Harris, S. E., Mundy, G. R., and Yoneda, T. (2000)

J. Bone Miner. Metab. 18, 132–13940. Saito, T., Kinoshita, A., Yoshiura, K., Makita, Y., Wakui, K., Honke, K.,

Niikawa, N., and Taniguchi, N. (2001) J. Biol. Chem. 276, 11469–1147241. Daluiski, A., Engstrand, T., Bahamonde, M. E., Gamer, L. W., Agius, E.,

Stevenson, S. L., Cox, K., Rosen, V., and Lyons, K. M. (2001) Nat. Genet. 27,84–88

42. Geiser, A. G., Zeng, Q. Q., Sato, M., Hirano, T., and Turner, C. H. (1998) Bone23, 87–93

43. Yamada, Y., Miyauchi, A., Goto, J., Takagi, Y., Okuizumi, H., Kanematsu, M.,Hase, M., Takai, H., Harada, A., and Ikeda, K. (1998) J. Bone Miner. Res.13, 1569–1576

44. Erlebacher, A., and Derynck, R. (1996) J. Cell Biol. 132, 195–21045. Filvaroff, E., Erlebacher, A., Ye, J., Gitelman, S. E., Lotz, J., Heillman, M., and

Derynck, R. (2002) Development 126, 4267–427946. Erlebacher, A., Filvaroff, E. H., Ye, J. Q., and Derynck, R. (1998) Mol. Biol. Cell

9, 1903–191847. Shull, M. M., Ormsby, I., Kier, A. B., Pawlowski, S., Diebold, R. J., Yin, M.,

Allen, R., Sidman, C., Proetzel, G., Calvin, D., Annunziata, N., andDoetschman, T. (1992) Nature 359, 693–699

48. Webb, D. J., Wen, J., Lysiak, J. J., Umans, L., Van Leuven, F., and Gonias,S. L. (1996) J. Biol. Chem. 271, 24982–24988

49. Akhoundi, C., Amiot, M., Auberger, P., Le Cam, A., and Rossi, B. (1994) J. Biol.Chem. 269, 15925–15930

50. Jilka, R. L., Hangoc, G., Girasole, G., Passeri, G., Williams, D. C., Abrams,J. S., Boyce, B., Broxmeyer, H., and Manolagas, S. C. (1992) Science 257,88–91

51. Ly, D. H., Lockhart, D. J., Lerner, R. A., and Schultz, P. G. (2000) Science 287,2486–2492

52. Border, W. A., and Ruoslahti, E. (1992) J. Clin. Invest. 90, 1–753. Shah, M., Foreman, D. M., and Ferguson, M. W. J. (1995) J. Cell Sci. 108,

985–100254. Cui, W., Fowlis, D. J., Bryson, S., Duffie, E., Ireland, H., Balmain, A., and

Akhurst, R. J. (1996) Cell 86, 531–54255. Ashcroft, G. S., Yang, X., Glick, A. B., Weinstein, M., Letterio, J. L., Mizel,

D. E., Anzano, M., Greenwell-Wild, T., Wahl, S. M., Deng, C., and Roberts,A. B. (1999) Nat. Cell Biol. 1, 260–266

3 C. J. Swallow, J. C. Macmillan, E. A. Partridge, T. T. Tajirian, K.Hay, L. Cindy, G. M. DiGuglielmo, K. Clemens, W. Jahnen-Dechent, M.Redston, J. L. Wrona, S. Gallinger, and J. W. Dennis, submitted forpublication.

�2-HS Glycoprotein and TGF-�-dependent Osteogenesis 19997