Juvenile dermatomyositis calcifications selectively displayed markers of bone formation




To determine the presence of small integrin-binding ligand N-linked glycoprotein (SIBLING) and bone components in juvenile dermatomyositis (DM) pathologic calcifications.


Calcifications were removed from 4 girls with juvenile DM symptoms for mean ± SD 36.9 ± 48.3 months and were stained for SIBLING proteins: full-length osteopontin (OPN), bone sialoprotein (BSP), dentin matrix protein 1 (DMP1), dentin phosphoprotein (DPP), and matrix extracellular phosphoglycoprotein (MEPE); bone markers: osteocalcin (OC), core-binding factor α 1 (CBFα1), and alkaline phosphatase (AP) for osteoblasts; tartrate-resistant acid phosphatase (TRAP) for osteoclasts; and the mineral regulators osteonectin (ON) and matrix Gla protein (MGP). The deposit center, periphery, adjacent connective tissue, and vascular endothelial cells were examined.


Alizarin red stained calcified deposits that did not localize with collagen, like bone, under polarized light. Hematoxylin and eosin stain revealed a paucity of connective tissue and absence of bone-like structures. The deposits, connective tissue, and vascular endothelial cells were positive for BSP, DPP, DMP1, and AP; MEPE was not detected. OC, ON, and MGP were present in the deposits and vascular endothelial cells; OPN and CBFα1 were present in deposits and connective tissue. TRAP-positive osteoclasts were localized to the calcification periphery.


The disorganized juvenile DM calcifications differ in structure, composition, and protein content from bone, suggesting that they may not form through an osteogenic pathway. Osteoclasts at the deposit surface represent an attempt to initiate its resolution.


Juvenile dermatomyositis (DM), the most common pediatric inflammatory myopathy, is a small-vessel systemic vasculopathy in which children present with symmetric proximal muscle weakness and a characteristic rash (1, 2). As many as 30% of patients with juvenile DM develop the painful complication of pathologic soft tissue calcifications (3). These calcifications are associated with chronic inflammation, usually occurring after a long period of untreated symptoms (4).

However, the pathologic soft tissue calcifications found in juvenile DM, although similar in composition to bone, are quite distinct. A previous study using Western Blot analysis identified osteopontin (OPN), osteonectin (ON), and bone sialoprotein (BSP) in juvenile DM calcifications (5). These proteins are also found in bone, but based on Fourier transform infrared spectroscopy, the juvenile DM calcifications exhibit a higher mineral to matrix ratio than bone (5), leading to the speculation that the mechanism of mineral deposition in juvenile DM might differ from that of bone formation. There may be other mediators of mineralization present in both bone and juvenile DM calcifications, since calcifications of soft tissue found in other diseases such as rheumatic valvular heart disease and scleroderma contain markers of bone formation (6, 7).

OPN and BSP both belong to the small integrin-binding ligand N-linked glycoprotein (SIBLING) protein family, along with fellow members: dentin matrix protein 1 (DMP1), dentinphosphoprotein (DPP), and matrix extracellular phosphoglycoprotein (MEPE). The SIBLING proteins play crucial roles in regulating bone and dentin formation, in addition to serving as cell signals. Their roles are governed by posttranslational modifications such as phosphorylation and glycosylation (8). Encoded on human chromosome 4, they contain an RGD integrin-binding domain that mediates cell–cell interactions (9). Other non-SIBLING mineralization mediators such as matrix Gla protein (MGP) and ON are also of interest, as they bind to calcium (10, 11). Previous studies found that juvenile DM patients with calcifications have a higher urinary MGP output (12), and identified MGP within the calcifications (13). Recent studies reported increased phosphorylated MGP within juvenile DM muscle from children with calcifications as compared with juvenile DM muscle from children without calcifications and normal controls (14). For the present study, osteoblast-specific markers core-binding factor α 1 (CBFα1), osteocalcin (OC), alkaline phosphatase (AP), and tartrate-resistant acid phosphatase (TRAP) activity were also used to identify osteoclasts. The proposed functions and tissue localization of these markers are shown in Table 1.

Table 1. Localization and function of bone-related protein antigens*
AntigenTissue localizationProposed function
  • *

    BSP = bone sialoprotein; OPN = osteopontin; DMP1 = dentin matrix protein 1; DSPP = dentin sialophosphoprotein; DPP = dentin phosphoprotein; MEPE = matrix extracellular phosphoglycoprotein; AP = alkaline phosphatase; CBFα1 = core-binding factor α 1; OC = osteocalcin; ON = osteonectin; MGP = matrix Gla protein.

BSP (LFMb-25)Bone, dentin, salivary glands, sweat glands, and kidneyMineral nucleator
Full-length OPN (LFMb-14)Bone, dentin, salivary glands, sweat glands, kidney inflammatory cells, and activated T cellsOsteoclast recruitment, mineralization inhibitor
DMP1 (LFMb-31)Bone, dentin, salivary glands, sweat glands, and kidneyMineral nucleator, mineralization regulator, regulator of osteocalcin
DSPP (LFMb-21; DPP domain CSRGDASYNSDESKDNG)Bone, dentin, salivary glands, sweat glands, and kidneyMineralization regulator, recruiter of inflammatory cells
MEPE (LFMb-33)Bone, dentin, salivary glands, sweat glands, and kidneyMineralization regulator
APCell surface of osteoblastsMarker of bone remodeling
CBFα1Osteoblast precursor cellsTranscription factor for osteoblast lineage
OCBone, osteoblastsMineralization regulator, hormone
ONBoneRegulate vascular homeostasis, tissue repair, collagen fibrillogenesis, mineralization regulator
MGPArteries, cartilageMineralization inhibitor


Patient population.

Four girls with probable juvenile DM, diagnosed by the attending physician (LMP) based on clinical criteria by Bohan and Peter (15), were enrolled in the study after obtaining age-appropriate informed consent (Institutional Review Board no. 2001-11530). Their demographic and clinical data are shown in Table 2.

Table 2. Demographic, clinical, and genetic data for 4 children with juvenile DM with removed calcinosis*
SampleSexAge at juvenile DM onset, yearsDuration of untreated disease, monthsAge at calcification removal, yearsLocationDAS muscleDAS skinTNFα-308DQA1*0501Nailfold capillary ERL/mm§
  • *

    DM = dermatomyositis; DAS = Disease Activity Score; TNFα-308 = tumor necrosis factor α 308; ERL = end row loop.

  • DAS for muscle strength and function; range 0–9, where 0 = normal and 9 = maximal impairment.

  • DAS for skin involvement severity and intensity; range 0–11, where 0 = normal and 11 = severe.

  • §

    Decreased ERL/mm in chronic juvenile DM with calcifications (normal value >7 ERL/mm).

1F2247.8Left patellar tendon03GA+3.6
2F2.13.412.2Elbow connective tissue10GA4.6
3F6.2108.318.5Elbow connective tissue17GG+3.7
4F1.11218.2Biceps muscle97GA+3.0
Mean ± SD 2.9 ± 2.336.9 ± 48.314.2 ± 5.2 2.8 ± 4.24.3 ± 3.4  3.7 ± 0.33

Clinical definitions.

The Disease Activity Score, a validated scoring of disease activity in skin and muscle (16), was determined by the same physician (LMP) during the clinical visit prior to the removal of the calcification. This score is a clinical estimate of disease activity that rates the active involvement of both skin and muscle for a total score of 20 points (16). The duration of untreated disease was defined as the interval of time between the first symptom (disease onset) and the date of diagnostic evaluation (4).

Genetic testing.

The primers and probes were obtained from Invitrogen (Madison, WI). Peripheral blood mononuclear cells were isolated from anticoagulated whole blood and stored at −80°C until DNA isolation was performed using a Puregene DNA isolation kit (Gentra Systems, Minneapolis, MN). The tumor necrosis factor α 308 (TNFα-308) polymorphism consists of a single-basepair substitution of an A for the more common G, and polymerase chain reaction was used to amplify a 107-bp fragment of that gene containing a region that incorporated a Nco I restriction site as previously described (17). Digestion with Nco I confirmed the genotype as GG, GA, or AA. Testing for DQA1*0501 was performed, as described elsewhere (18).

Sample collection.

Four girls requested surgical removal of painful calcifications. The samples were fixed overnight in 10% formalin (pH 7.0). Due to the difficulty of sectioning calcified tissue, patient sample 4 was extensively decalcified (10% hydrochloric acid). All tissues were paraffin embedded and sectioned at 4 μm (Histology Department, Children's Memorial Hospital, Chicago, Illinois).


Standard techniques were followed to deparaffinize, rehydrate, and retrieve antigens with citrate buffer (pH 6.0) for 20 minutes at 95°C (19). The sections were incubated with the following primary antibodies at the specified dilutions: CBFα1 1:25, OC 1:10 (R&D Systems, Minneapolis, MN), ON 1:10 (Developmental Studies Hybridoma Bank, Iowa City, IA), AP 1:50 (Sigma-Aldrich, St. Louis, MO), BSP 1:100, DMP1 1:50, DPP 1:400, MEPE 1:300, full-length OPN 1:200 (provided by Larry W. Fisher, PhD, National Institute of Dental and Craniofacial Research, National Institutes of Health), and MGP 1:50 (Proteintech Group, Chicago, IL). The primary antibody was detected by standard methodology (20). Substitution of the primary antibody with mouse IgG control, rabbit IgG control (catalog nos. 08-6599, 08-6199; Zymed, South San Francisco, CA), and rat IgG control (catalog no. R2b00, Invitrogen) served as negative controls. Human osteosarcoma (MG-63) cell lines (American Type Culture Collection, Manassas, VA) and paraffin-embedded human iliac crest sections served as marker-positive controls.


The sections were stained with hematoxylin and eosin (H&E) (21), alizarin red (22), and TRAP (23) by standard methodology.

Image acquisition.

Serial sections of each patient sample were examined, and areas of calcified tissue (confirmed by alizarin red staining) were photographed under the same conditions. Images of stained sections were acquired using Openlab computer software, version 4.04 (Improvision, Lexington, MA), and a Leica DMR-HC microscope (Leica, Wetzlar, Germany), coupled to a QImaging Retiga 4000R camera (QImaging, Surrey, British Columbia, Canada) and edited using Adobe Photoshop CS2 software (Adobe Systems, San Jose, CA).

Data analysis.

Images of serial sections stained for each marker were examined by 2 independent observers. Positive staining was defined as areas of dark brown comparable to positive control slides. The presence or absence of each marker within the center or periphery (outer edge) of calcium deposits, surrounding connective tissue, and endothelial cells was recorded as the number of positive samples over the total number of samples tested. For each marker, a tissue location was defined as positive staining if at least 2 of 3 calcified samples stained showed localized antigen-specific antibody.


H&E and alizarin red staining.

Calcium deposits were confirmed by alizarin red staining. No cellular or bone-like trabecular structures were observed within the deposits (Figure 1).

Figure 1.

Alizarin red stain, tartrate-resistant acid phosphatase (TRAP) stain, hematoxylin and eosin (H&E) stain, and immunohistologic stain for dentin phosphoprotein (DPP), osteopontin (OPN), matrix extracellular phosphoglycoprotein (MEPE), bone sialoprotein (BSP), dentin matrix protein 1 (DMP1), alkaline phosphatase (ALP), core-binding factor α 1 (Cbfα1), osteonectin (ON), osteocalcin (OC), and matrix Gla protein (MGP) in mineral deposits and adjacent connective tissue (magnification × 10). Brown indicates positive stain. A, magnified region in the other panels, B, indicated region of TRAP-positive osteoclasts at magnification × 10, C, human bone as positive control for MEPE, D, positive control for BSP, E, positive control for ALP, F, representative region of surrounding cells demonstrating DMP1 reactivity within the cytoplasm (magnification × 63). Arrows indicate collagen. * Picture taken at magnification × 1; ** Picture taken at different site. NC1 = negative control for DPP, OPN, MEPE, BSP, DMP1, ALP, ON, and OC; NC2 = negative control for Cbfa1; NC3 = negative control for MGP; PL mineral = juvenile dermatomyositis mineral deposits under polarized light; PL bone = normal bone taken under polarized light.

Immunohistochemical staining.

Typical results for the immunostaining of juvenile DM calcification samples are shown in Figures 1 and 2, whereas a summary of the results from patient samples for full-length OPN, BSP, DMP1, DPP, MEPE, AP, OC, CBFα1, ON, and MGP are shown in Table 3, and are shown as a schematic representation in Figure 3. The extensively decalcified sample was not included in the final tallies in Table 3, and the presence of the markers in this sample is noted by an asterisk in Figure 3.

Figure 2.

Alizarin red stain, tartrate-resistant acid phosphatase (TRAP) stain, hematoxylin and eosin (H&E) stain, and immunohistologic stain for dentin phosphoprotein (DPP), osteopontin (OPN), matrix extracellular phosphoglycoprotein (MEPE), bone sialoprotein (BSP), dentin matrix protein 1 (DMP1), alkaline phosphatase (ALP), core-binding factor α 1 (Cbfα1), osteonectin (ON), osteocalcin (OC), and matrix Gla protein (MGP) in vascular endothelial cells (magnification × 40). Brown indicates positive stain. NC1 = negative control for DPP, OPN, MEPE, BSP, DMP1, ALP, ON, and OC; NC2 = negative control for Cbfa1; NC3 = negative control for MGP.

Table 3. Positive samples out of number examined for each antigen/marker*
 Vascular endothelial cellsCalcification centerCalcification peripheryAdjacent tissue
  • *

    TRAP = tartrate-resistant acid phosphatase; see Table 1 for additional definitions.

Alizarin red0/33/33/30/3
Figure 3.

Schematic representation of the composition of a juvenile dermatomyositis calcification. * Presence of marker in both calcified and decalcified sample. DMP1 = dentin matrix protein 1; OC = osteocalcin; ON = osteonectin; MGP = matrix Gla protein; DPP = dentin phosphoprotein; BSP = bone sialoprotein; ALP = alkaline phosphatase; OPN = osteopontin; Cbfα1= core-binding factor α 1.

Calcium deposits within 3 samples stained dark red with alizarin red. In contrast, sample 4, which was extensively decalcified prior to sectioning, did not exhibit the dark red staining typical of alizarin red. TRAP staining revealed the presence of osteoclasts (stained dark pink) around the periphery of mineral deposits in all 3 calcified samples, but not in the decalcified one. Figure 1F is a representative picture of all of the stains taken from the DMP1 stain, which demonstrates that the pattern of reactivity within the cells surrounding the calcifications and vascular endothelial cells (Figure 2) is primarily cytoplasmic in nature. After inspecting H&E sections of juvenile DM calcification samples under polarized light (Figure 1), we identified the total absence of embedded collagen fiber within mineral deposits as compared with sections of normal trabecular bone, despite its presence in other parts of the tissue sample.

OPN in full-length form is present within the center and periphery of mineral deposits in 3 juvenile DM calcification samples, as well as in the decalcified sample. DMP1, MGP, OC, and ON were identified both in the center and in the periphery of mineral deposits in the 3 calcified samples, whereas these molecules were not present within the decalcified sample deposits. DPP was observed within the center of mineral deposits in 3 calcified samples and in the periphery of 2 calcified samples, as well as in the decalcified sample. BSP was observed in the center of 1 calcification sample and in the periphery of 2 calcification samples, but not in the decalcified sample. AP-positive cells were found on the periphery of mineral deposits in the 3 calcified samples, but only in the center of mineral deposits in 1 of the samples. CBFα1 was present in the center and periphery of mineral deposits in 2 of the samples. The site of the mineral deposits in the decalcified sample did not exhibit AP or CBFα1 in the center or periphery.

With respect to staining of vascular endothelial cells, BSP, DMP1, and OC were present in all 3 calcified samples, as well as in the decalcified sample. DPP, MGP, ON, and AP were expressed in 2 calcified samples, and each of these markers, with the exception of ON, was present in the decalcified sample. OPN and CBFα1 were observed in only 1 calcified sample.

In the adjacent connective tissue, BSP and DPP were observed in 3 calcified samples, as well as in the decalcified sample. OPN, DMP1, and AP were expressed in 2 calcified samples, as well as in the decalcified sample. CBFα1 was also present in 2 calcified samples, but not in the decalcified sample. ON was also detected in 1 calcified sample, but not in the decalcified sample. Neither MGP nor OC was observed in the connective tissue of any of the samples. The presence of MEPE was not observed in any tissues analyzed.

Mouse isotype controls (for DPP, OPN, MEPE, BSP, DMP1, AP, ON, and OC) and rabbit isotype controls (for MGP) were negative in all tissues, whereas rat isotype controls (for CBFα1) showed slight positive staining within mineral deposits.


To our knowledge, this is the first immunohistochemical study of juvenile DM calcifications assessing the mineralized tissue for the presence of SIBLING proteins OPN, BSP, DMP1, and DPP; non-SIBLING mineralization regulators ON and MGP; osteoblast-specific markers OC, AP, and CBFα1; and osteoclast-specific TRAP.

SIBLING proteins initially identified in bone and dentin (8) were then demonstrated in specific cancers in which calcified tissue occurred, such as breast and thyroid carcinomas (24). Additionally, osteoblast-like phenotypes have been observed within calcified tissue in cases of rheumatic valvular heart disease and scleroderma (6, 7), which first led us to investigate the presence and location of bone-forming and reabsorbing cells within juvenile DM calcifications.

OPN has been proven to be an important mineralization inhibitor and has aspartic acid–rich calcium-binding domains (25), as well as extensive phosphorylation, which may account for its accumulation and localization at calcium deposits. OPN has also been shown to play a role in recruitment and adhesion of osteoclasts (26), which could be associated with the positive TRAP staining for osteoclasts around the periphery of mineral deposits. Furthermore, macrophages, which are present in the inflammatory infiltrate in juvenile DM muscle (27), produce OPN, which may opsonize the calcifications, thus serving as an attractant for other macrophages and phagocytic cells interacting via the RGD-binding domain (28).

Macrophages migrating into the injury site could also differentiate into osteoclasts, under the influence of the abundant chemokines at the inflammation sites. Previous studies have shown that biopsy samples from juvenile DM patients with chronic muscle weakness exhibit elevated interleukin 1 (IL-1) expression, and juvenile DM patients positive for the TNFα-308 A allele exhibit higher TNFα expression in muscle compared with those patients without this allele (29, 30). Both IL-1 and TNFα have been proven to be potent initiators of osteoclast formation (31). In the present study, osteoclasts were identified by TRAP staining at the outer edge of calcium deposits. The larger deposits seemed to attract more osteoclasts than smaller deposits. Nevertheless, the persistence of calcifications despite the presence of osteoclasts indicates that the mineral reabsorbing activity of these cells appears to be neither sufficient nor effective in dissolving the deposited mineral.

DMP1, BSP, and DPP all showed similar patterns of staining within endothelial cells and adjacent connective tissue in both the calcified and decalcified samples; however, those proteins were only detected in the mineral deposits of the calcified samples and not in the decalcified sample. The presence of BSP suggests that it may play a role in osteoclast differentiation (32), despite in vitro studies that demonstrated that BSP expression fosters osteoblast differentiation, leading to increased BSP expression and mineralization (33). DMP1, a transcriptional signal occurring early during differentiation of osteoblasts, initiates mineralization during the final steps of osteoblast differentiation, and is a regulator of the osteoblast gene OC (8, 34, 35). DMP1 binds calcium, and is a nucleator of hydroxyapatite, the mineral that is present in high concentrations in juvenile DM calcification (36). The centralized location of DMP1 within mineral deposits is consistent with this function. DPP is an effective stimulatory molecule during the dissolution of the dentin matrix by attracting inflammatory neutrophils and stimulating the release of cytokines such as IL-1β and TNFα by macrophages (37, 38). The presence of DPP within calcifications, soft tissue, and vascular endothelial cells may serve as an effective attractant for macrophages in an effort to disperse the calcifications, consequently recruiting additional lymphocytes and exacerbating inflammation.

Although MEPE is a component of bone, its role in mineralization is debatable. MEPE inhibits mineralization in some studies (39), while in others it has been shown to promote bone regeneration (40). The absence of MEPE in all of the juvenile DM calcifications further suggests that the mechanism of calcification in juvenile DM differs from that of bone.

ON, one of the most highly expressed noncollagenous proteins in bone, binds calcium, and has been identified in tissues undergoing repair or remodeling, as well as in malignant tumors (10, 41). ON also regulates vascular homeostasis by interacting with platelet-derived growth factor, fibroblast growth factor 2, and vascular endothelial growth factor (10). Expression of ON within the mineral deposits may indicate a tissue repair mechanism that is insufficient to maintain homeostasis, thus contributing to deposition of mineral deposits.

One of the few osteoblast-specific proteins, OC is secreted by osteoblasts into the circulation and is notably regulated via the transcription factor CBFα1 (42), which we identified in the juvenile DM deposits. Additionally, OC-deficient mice exhibit increased bone formation, as well as higher bone mass with improved function and defective remodeling (43, 44).

AP activity detected within vascular endothelial cells, adjacent tissue, and on the periphery of calcifications suggests an osteoblast-like mechanism of mineralization, and the presence of OC, CBFα1, and AP within juvenile DM calcifications suggests that osteoblast-like cells may be present; however, the structure of the mineral deposits does not exhibit the morphology of physiologic bone often seen in rheumatic valvular heart disease (7).

In the past, investigators identified MGP as one of the proteins present in juvenile DM calcifications (13), and recent studies have shown phosphorylated MGP to be present within diseased juvenile DM muscle from patients with calcifications as compared with controls (14). However, immunohistologic examination to localize MGP within juvenile DM calcifications has not been reported. The present study documented that MGP was involved in both the mineral deposits and endothelial cells in all of the calcified samples, but it was not detected in the decalcified sample. This observation suggests that MGP was bound to the precipitated mineral, and may be involved at the initiation stage of the calcification process. MGP may possess a function similar to some of the large calcium-binding complex proteins, such as fetuin A and OPN, which, along with MGP, have been reported to form calciprotein particles with calcium phosphate in the circulation to prevent the deposition of minerals (45). Fetuin appears to be increased in young children (46), and may contribute to the increased frequency of calcifications in this group of children with prolonged untreated inflammation central to juvenile DM.

Juvenile DM is characterized as a small-vessel systemic vasculopathy. We have previously shown that a long duration of untreated disease was accompanied by a marked decrease in nailfold capillary end row loops, which did not revert to normal with disease control in many cases (47). For this reason, we examined the vasculature for the presence of SIBLING and bone markers. From the H&E stain in Figure 2, we observed that the vascular endothelial cells are activated and colocalize with the presence of the observed markers within the samples. To our knowledge, there have not been any studies investigating the expression of these proteins in the vasculature of idiopathic calcinosis. However, the presence of BSP and OPN has been identified within the vasculature associated with prostate and breast tumors where calcified tissue occurs. Both BSP and OPN have been proven to be proangiogenic factors and interact via the RGD binding domain with the αvβ3 receptor on activated vascular endothelial cells in tumor vasculature (24). Endothelial cells have also been reported to express ON (48) and MGP (49). We have confirmed the previous reports and further identified the expression of DMP1, DPP, AP, and OC in endothelial cells from mineralized tissue. DMP1 and DPP may exhibit functions that are similar to those of BSP and OPN, since they also contain the RGD-binding domain. The presence of CBFα1 in the surrounding tissue indicates osteoblast progenitor cells, whereas observation of the expression of OC and AP in vascular endothelial cells further identifies osteoblast-like cells in the surrounding tissues and vasculature. These observations reveal the interplay between immune regulation, angiogenesis, and calcification, but further studies are needed to determine the mechanism by which these markers operate in calcifications that occur as a consequence of the chronic inflammation present in juvenile DM.

Previous studies from our laboratory demonstrated an increased number of apoptotic cells within inflamed juvenile DM muscle as compared with control age-matched pediatric muscle tissue (20). Studies of the juvenile DM calcification by electron microscopy have documented damaged mitochondria (Zhao: unpublished observations) in the surrounding soft tissue, which may release internal calcium stores, thus providing the nidus for mineral deposition. Examination of the deposits with polarized light clearly shows that deposition of mineral is not spatially associated with collagen, as is present in normal trabecular bone. After identifying the SIBLING and bone formation–associated proteins within juvenile DM calcifications, we are now able to further explore the possible mechanisms by which these proteins operate in juvenile DM calcifications.

In conclusion, we have presented evidence for the presence of osteoclasts and osteoblast markers in the absence of observable osteoblasts. Osteoclasts at the surface of the juvenile DM deposit may be present in an attempt to resolve the calcification. Mineral-binding proteins may attach to the deposits after they form, but the presence of osteoblast-specific markers (CBFα1, OC, and AP) indicates that these cells may also be involved. Since the juvenile DM calcifications differ from bone in structure, composition, and protein content, we speculate that they may not be formed via a classical osteogenic pathway.


Dr. Pachman had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Urganus, Zhao, Pachman.

Acquisition of data. Urganus, Zhao.

Analysis and interpretation of data. Urganus, Zhao, Pachman.

Manuscript preparation. Urganus, Zhao, Pachman.

Statistical analysis. Urganus, Zhao.


We are grateful for the comments given by Dr. Adele L. Boskey, the technical assistance from the core center (ALB), and the antibodies provided by Dr. Larry W. Fisher.