Increased Marrow-Derived Osteoprogenitor Cells and Endosteal Bone Formation in Mice Lacking Thrombospondin 2


  • Presented in part in preliminary form as a short communication (Trans Orthop Res Soc 1999;302).


The phenotype of thrombospondin 2 (TSP2)–null mice includes abnormalities in collagen fibrils and increases in ligamentous laxity, vascular density, and bleeding time. In this study, analyses by computerized tomography (CT) revealed that cortical density was increased in long bones of TSP2-null mice. Histomorphometric analysis showed that the mid-diaphyseal endosteal bone formation rate (BFR) of TSP2-null mice was increased in comparison with that of wild-type (WT) animals. Although microgeometric analysis showed that periosteal and endosteal radii were reduced, the mechanical properties of femurs from TSP2-null mice were not significantly different from those of controls, presumably because of the concomitant increase in endosteal bone mass. Bone loss in ovariectomized mice was equivalent for WT and mutant mice, a finding that indicates that TSP2-null animals are capable of normal bone resorption. To further explore the cellular basis for the increased endosteal BFR in TSP2-null mice, marrow stromal cells (MSCs) were isolated and examined in vitro. These cells were found to be present in increased numbers in a colony forming unit (CFU) assay and showed an increased rate of proliferation in vitro. We conclude that TSP2 regulates the proliferation of osteoblast progenitors, directly or indirectly, and that in its absence endosteal bone formation is increased. (J Bone Miner Res 2000;15:851–862)


Thrombospondin 2 (TSP2) is an extracellular protein that is distributed broadly among connective tissues in the developing mouse.(1–3) Mice with a targeted disruption of the Thbs2 gene (TSP2-null mice) have a number of abnormalities, including alterations in the size and shape of dermal and tendon collagen fibrils, increased cortical density of long bones, increased vascular density, and a bleeding diathesis.(4) TSP2-null skin fibroblasts show decreased adherence in vitro and produce increased levels of matrix metalloproteinases.(5) As a consequence of these abnormalities, increased neovascularization in response to subdermal sili-cone rubber implants(6) and an accelerated healing of excisional full-thickness dermal wounds(7) have been observed.

The functions of both TSP1 and TSP2, as discerned by studies of the phenotypes of mice that lack these proteins,(4,8,9) support the notion that these proteins are members of a group of extracellular molecules that have been termed matricellular proteins.(10) Matricellular proteins do not subserve primarily structural roles in the extracellular matrix (ECM) but function contextually as adapters and modulators of cell-matrix interactions. These proteins influence cellular properties such as adhesion, migration, growth, and differentiation by their ability to interact with multiple cell-surface receptors, as well as with cytokines, proteases, and the ECM.(10)

Although TSP2-null mice have increased cortical bone density,(4) the function of TSP2 in bone remains unknown. TSP2 is present in the developing murine skeleton,(3) and is synthesized by both transformed osteoblasts(11) and primary mouse osteoblast precursors (K. Hankenson, 1999, unpublished data). TSP1, a paralog with the same domain structure as TSP2 and with considerable sequence identity,(12) is also present in the developing skeleton(1) and in bone(13) and is produced by osteoblasts in culture.(14) Despite its presence in the developing and adult skeleton, no bone abnormalities have been reported in TSP1-null mice.(8,9) TSP1, and possibly TSP2, serve as ligands for the integrin receptor αvβ3,(15,16) cell-surface-associated heparan sulfate proteoglycans,(17,18) CD36,(19) and the integrin-associated protein (CD47).(20) By interacting directly with these receptors on bone cells, TSP2 could regulate bone density by decreasing osteoblast and/or increasing osteoclast activity. Alternatively, TSP2 could function in a more indirect fashion in bone by binding growth factors, thereby modulating their cellular effects, or by modulating bone vascularity. TSP1 binds and activates transforming growth factor β1 (TGF-β1)(21,22) and also binds basic fibroblast growth factor (bFGF) and platelet-derived growth factor (PDGF),(23,24) and both TSP1 and TSP2 inhibit neovascularization in vitro and in vivo(25)

The mutation or targeted disruption of a number of genes expressed in bone, including src,(26) osteoprotegerin,(27) macrophage colony-stimulating factor,(28) and cathepsin K,(29) results in reduced osteoclast formation or function, which leads in turn to increased bone density, a condition termed osteopetrosis. These bone-modeling disorders are characterized by decreases in the absorption of primary spongiosa and increases in trabecular bone volume. TSP2-null mice do not have any of the common morphological characteristics of osteopetrosis,(4) a finding that suggests that a defect in bone resorption does not lead to increased bone density in TSP2-null mice. Rather, increased bone density in TSP2-null mice may develop from increased bone formation. Spontaneous or targeted mutations that lead to increases in bone formation are uncommon; however, mice with a disruption of the osteocalcin gene do have increased bone formation(30) and represent a notable exception.

To better define the process responsible for increased bone density in TSP2-null mice, we have conducted a comprehensive study of TSP2-null bone. Our findings indicate that increased bone density is a result of a preferential increase in endosteal bone production, but that these mice are fully capable of bone resorption. Further, we have shown that increased bone formation is associated with an increased generation of osteoblast precursors (marrow stromal cells [MSCs]) in bone marrow. Increases in MSC numbers have been correlated previously with increases in bone formation in both the rat(31) and the mouse;(32,33) however, this is the first report of the regulation of MSC proliferation and, in turn, endosteal bone modeling by a component of the ECM.


Bone labeling, collection, and histomorphometry

Two groups of age- and sex-matched TSP2-null and wild-type (WT) mice (group 1, five WT and five TSP2-null 4-month-old females; group 2, six WT and six TSP2-null 6-month-old males) received intraperitoneal (ip) injections of calcein (10 μg/g body weight) twice daily on day 0 and again on day 7. Mce were killed humanely on day 8 and femurs were fixed in cold, 70% ethanol. After dehydration, bone samples were infiltrated with a methyl methacrylate-dibutyl phthalate plastic composite. Cross-sections of the femoral midshaft, approximately 200 μm thick, were cut on a low-speed diamond wheel saw and ground and polished to a thickness of approximately 30 μm (± 10 μm). Static and dynamic histomorphometry of cortical bone was performed using an OsteoMeasure software program (OsteoMetrics, Inc., Atlanta, GA, U.S.A.) interfaced with a Nikon Eclipse E400 light/epifluorescent microscope and video subsystem (Nikon, Inc., Melville, NY. U.S.A.) on single, 30-μm unstained midshaft sections. Cross-sectional areas and periosteal and endosteal bone formation rates (BFRs) were determined. Total bone and marrow cavity areas were measured and the total cortical bone area was calculated by subtracting marrow from total bone area. To determine dynamic parameters, single-labeled calcein perimeter, double-labeled calcein perimeter, and interlabel width were measured on both the periosteal and the endosteal surfaces. The mineralizing surfaces, mineral apposition rates, and the BFRs were then calculated for the two surfaces.

Peripheral quantitative computerized tomography

A Norland/Stratec XCT-RM instrument (Norland Medical Systems, White Plains, NY, U.S.A.) was used to perform peripheral quantitative computerized tomography (pQCT) scans of femurs from the same 12 6-month-old males that were utilized for histomorphometry. The metaphyseal region of the proximal femur was positioned for scanning at a site that was equidistant between the proximal articular surface and the midpoint of the diaphysis. The position was verified using scout views and two 0.5-mm slices perpendicular to the long axis of the femoral shaft were collected. A single 0.5-mm scan of the femoral midshaft also was obtained. All scans were analyzed using a threshold for delineation of external boundary, as well as an area peel for subdivision into a cortical/subcortical region and a cancellous subregion. The bone mineral content (g/cm3), bone mineral density (g/cm3), and areas of each subregion were then determined by system software.

Microcomputed tomography

The geometry of five WT and five TSP2-null femurs from 4-month-old, female mice was analyzed using a microcomputed tomography (μCT) system.(34) This system provides a complete three-dimensional digitization of the structure with a resolution of 20 μm in the x (anterior-posterior), y (medial-lateral), and z (axial) directions. Images were thresholded to distinguish bone from nonbone voxels using a custom edge-detection algorithm. The analysis region of each femur was defined as the mid-50% of the whole bone length. This analysis region was then further divided into three segments; segment 1 (proximal), segment 2 (middiaphysis), and segment 3 (distal) were determined as 40%, 30%, and 30% of the analysis region, respectively (Fig. 1). Segment 1 represents an area of large muscle attachment, segment 2 represents the middiaphyseal region, and segment 3 contains the initial metaphyseal flare. Cortical thickness, cross-sectional area, and moments of inertia (Ixx, Iyy, and J) were calculated for each slice within a segment and then averaged along the entire length of that segment. Moments of inertia characterize the distribution of bone material from the neutral bending axis of the bone and are related directly to mechanical properties in bending. In addition, cortical thickness, inner fiber length, and outer fiber length of each slice were calculated for the femurs at 10° intervals around the circumference of the bone and averaged for the entire segment. Inner and outer fiber lengths measure the distance from the centroid of the slice to the endosteal and periosteal surfaces, respectively, and reflect alterations in the endosteal and periosteal perimeters.

Figure Fig. 1..

Three-dimensional μCT reconstruction of whole femurs from TSP2-null and WT mice. Images show regional decreases in inner and outer fiber lengths (corresponding to decreased endosteal and periosteal perimeters, respectively), as well as increases in cortical thickness in TSP2-null femurs.

Mechanical testing

Whole bone mechanical properties were determined using the same femurs (five WT and five TSP2-null femurs from 4-month-old female mice) studied by μCT by testing to failure in four-point bending as previously described.(35) Specimens were oriented with their anterior surfaces facing upward, and the middiaphysis was loaded using an MTS servohydraulic testing machine (MTS Systems Corp., Eden Prairie, Minneapolis, MN, U.S.A.) at a constant displacement rate of 0.5 mm/s. Load-displacement data were acquired using LabView (National Instruments, Austin, TX, U.S.A.) software, and loads and displacement to yield and failure were measured directly. Stiffness was calculated as the slope of the linear portion of the load-displacement curve.


Five TSP2-null and five WT female mice, 5 months of age, were anesthetized with xylazine and ketamine by ip injection and were ovariectomized using a routine surgical procedure.(36) Four TSP2-null and five WT mice served as sham controls. Four weeks after ovariectomy, femurs were harvested and evaluated using pQCT.


Serum was collected from mice at the time of death. Osteocalcin levels were determined in duplicate using a commercially available radioimmunoassay (RIA) kit (BTI, Stoughton, MA, U.S.A.)

MSC culture

One-month-old mice were weighed and killed by rapid cervical dislocation. Both left and right tibia and femur complexes were isolated aseptically and harvested into MSC media (α-minimum Eagle's media and 10% fetal calf serum, containing 100 IU/ml penicillin, 100 μg/ml streptomycin, 10 μM amphotericin-B, and 50 μg/ml sodium ascorbate). The majority of adherent soft tissues and the metaphyses were removed under sterile conditions. Whole marrow was expelled from the bones with several flushes through a 23-gauge needle. Single-cell marrow suspensions were made by aspiration of the marrow preparation through a 23-gauge needle for 3 minutes. Cells were pelleted, resuspended, and counted using a hemocytometer to determine the total number of marrow cells obtained from both tibias and femurs. Appropriate cell suspensions were made for subsequent cell plating.

MSC colony-forming unit fibroblastic assay

Marrow cells were plated in duplicate 60-mm dishes at a density of 1 × 105 cells/cm2. To maintain the influence of nonadherent cells, only 1/3 of the volume of medium was removed and replaced on day 3 and day 6 and the plates were harvested on day 9. At harvest, plates were rinsed with phosphate-buffered saline (PBS), fixed with citrate buffer, and stained and counterstained for alkaline phosphatase (AP) activity using a commercially available kit (Sigma, St. Louis, MO, U.S.A.). Colonies of cells containing greater than 20 cells were counted manually and the number of AP-positive colonies (containing a cell population of at least 20% AP-positive cells) was determined in a similar manner.

To determine colony size, WT and TSP2-null colony-forming unit fibroblastic (CFU-F) plates were digitally scanned at 100% magnification using an Arcus II scanner (Agfa-Gevaert N.V., Mortsel, Belgium). Using ImageQuant (IQ) software (Molecular Dynamics, Sunnyvale, CA, U.S.A.), colonies were thresholded to exclude those that were less than 1.5 mm2 and delineated using the SpotFinder function. The relative area of each colony (expressed in IQ units) was determined and converted to absolute area (mm2) by comparing IQ units to a standard curve of IQ units versus actual area (mm2).

Proliferation assay

Isolated MSCs were plated at high density and grown to confluence over 12 days. Confluent cells were harvested by trypsinization and resuspended in MSC medium. Cells were pelleted by centrifugation and resuspended as single-cell suspensions by passing through a 23-gauge needle for 3 minutes. Cells were counted with a hemocytometer. Cell dilutions of 1 × 105 cells/ml were made for each experimental group and 2.0 ml of diluted cells were plated in quadruplicate into each well of a 6-well plate. For two of the wells for each sample, medium was changed to MSC medium containing 10 mM β-glycerol phosphate (BGP) and 1 × 10–8 M dexamethasone (Dex) to induce differentiation to the osteoblast phenotype. On day 12, the cells were starved in serum-free medium for 24 h and then changed back to medium with serum for 16 h before labeling for 5.5 h in complete medium with 5 μCi of 3H[thymidine] (Amersham Pharmacia Biotech, Piscataway, NJ, U.S.A.). Cells were harvested by rinsing with PBS and then mixed with a detergent solution (1.0 M NH4OH/0.2% Triton X-100) to lyse the cells. A portion of the cell lysate was precipitated with 25% trichloroacetic acid (TCA) and the acid-insoluble precipitate was pelleted by centrifugation and rinsed once with 10% TCA. TCA-precipitated material was resuspended in 0.5 M NaOH and radioactivity was determined using a scintillation counter. Counts per minute (CPMs) were normalized to DNA content, as determined by A260.

Statistical analysis

Results are presented as mean and standard deviation. Data were analyzed for statistical significance using one-way analysis of variance (ANOVA) procedures. The analysis of correlation between serum osteocalcin and CFU-F was performed using a linear regression to determine R and p values. Unless otherwise indicated, results are considered to be statistically significant if p < 0.05.


Dynamic histomorphometry and pQCT

Previously, reported experiments had shown an increase in mid-diaphyseal total bone density and cortical thickness in TSP2-null mice.(4) Despite the difference in diaphyseal cortical bone, trabecular density was equivalent between WT and TSP2-null mice and TSP2-null trabeculae were histologically normal. To study the role of cortical bone formation in the generation of increased bone density, 4-month-old female mice were double-labeled with calcein 1 week before death and bone harvest. Histological examination of midfemoral cross-sections revealed an increase in the separation and in the length of the endosteal calcein double-label (Fig. 2). Histomorphometric analyses showed that TSP2-null mice have a 23% decrease in marrow area and an 11% decrease in endosteal perimeter (Table 1). These geometric alterations develop because TSP2-null mice have a 2- to 3-fold increase in endosteal BFR and a 58% increase in endosteal mineral apposition rate. TSP2-null mice also have a 2-fold increase in endosteal mineralizing surface. Interestingly, the difference in endosteal dynamic parameters does not extend to the periosteal surface. Periosteal parameters for TSP2-null mice appear to be decreased relative to the controls, for example, the 17% decrease in periosteal BFR, but the differences do not achieve statistical significance.

To study this unusual bone phenotype further, 6-month-old male mice were evaluated by pQCT. These mice also displayed statistically significant alterations in bone geometry, characterized by a 13% increase in total bone density and a 6% increase in cortical density (Table 2). Additionally, there was a trend toward decreased periosteal and endosteal circumferences and increased cortical area and cortical thickness; however, these differences did not reach statistical significance. Mice that were evaluated by pQCT were also double-labeled with calcein and showed a 69% increase in endosteal BFR and a 57% increase in endosteal mineralizing surface (Table 3). The other histomorphometric parameters were not significantly different; however, the trends that were noted for 4-month-old females (Table 1) also were observed for 6-month old male mice. These trends included a decrease in marrow area and endosteal perimeter and decreases in periosteal perimeter, mineralizing surface, and BFR.

Figure Fig. 2..

Increased endosteal bone formation as judged by calcein double-label of TSP2-null long bone. Femurs of 4-month-old female mice that were used for evaluation by pQCT also were analyzed using fluorescence microscopy. The arrow indicates the double-labeled calcein surface in these femoral cross-sections. Notice that both the separation and the length of the calcein double-label are greater for the TSP2-null cross-section (magnification, × 40).

Table Table 1.. Increased Endosteal Bone Formationin TSP2-Null Micea
Morphometric parameterMean (SD)Mean (SD)
  1. MAR, mineral apposition rate; MS, mineralizing surface.

  2. a Four-month-old females.

  3. * Indicates that TSP2-null and WT are significantly different at p < 0.05; n = 5.

Marrow area (mm2)0.84 (0.08)0.65 (0.05)*
Endosteal perimeter (mm)3.41 (0.21)3.02 (0.15)*
Endosteal MS (%)19.84 (11.23)43.38 (8.00)*
Endosteal MAR (μ/day)4.04 (1.35)6.40 (0.84)*
Endosteal BFR (μ32 per day)1.25 (0.88)3.12 (0.92)*
Periosteal perimeter (mm)4.89 (0.41)4.57 (0.14)
Periosteal MS (%)47.56 (22.44)36.91 (9.93)
Periosteal MAR (μ/day)4.76 (0.90)5.60 (0.65)
Periosteal BFR (μ32 per day)2.68 (1.47)2.28 (0.72)
Table Table 2.. Increased Bone Densityin TSP2-Null Micea
pQCT parameterMean (SD)Mean (SD)
  1. a Six-month-old males.

  2. * p < 0.05; n = 6.

Periosteal circumference (mm)4.86 (0.29)4.74 (0.49)
Endosteal circumference (mm)3.16 (0.35)2.83 (0.38)
Cortical area (cm2)1.11 (0.11)1.14 (0.18)
Cortical density (mg/cm3)681.53 (15.77)725.35 (26.00)*
Cortical thickness (mm)0.279 (0.174)0.304 (0.028)
Total density (mg/cm3)464.73 (27.93)523.13 (52.37)*

Microgeometric characterization and mechanical properties

To further define the geometric alterations associated with the TSP2-null phenotype, μCT was utilized to study the bones of 4-month-old female mice. Analysis of the three-dimensional reconstructed images of μCT scans (Fig. 1) revealed obvious geometric differences between WT and TSP2-null mice. Quantitatively, there was a statistically significant (p < 0.05) increase in cortical thickness in segment 1 of the knockout mice when compared with WT controls (TSP2-null = 0.41 mm; WT = 0.37 mm). There was a similar ∼ 10% increase in cortical thickness in segment 2 (TSP2-null = 0.34 mm; WT = 0.30 mm), but the values were not significantly different. More refined analysis of the three segments at 10° increments around the circumference of each segment revealed significant quantitative shape differences between TSP2-null and WT specimens. Proximal (segment 1) specimens from knockout animals showed significantly decreased outer fiber lengths (defined as the radius extending from the cross-sectional marrow center to the periosteal surface) in anteromedial and posterolateral regions (Fig. 3A). Additionally, inner fiber length (defined as the radius extending from the cross-sectional marrow center to the endosteal surface) was decreased in specimens from knockout animals for almost 3/4 of the bone circumference (20–280°; Fig. 3B). Cortical thickness was significantly increased in the knockout animals in the medial, posterolateral, and anterolateral regions (Fig. 3C). Similar results were found for segments 2 (mid-diaphysis) and 3 (distal; results not shown), although the decrease in inner fiber length in specimens from the TSP2- null mice was even more substantial. These results suggest that altered bone formation does not occur evenly over the entire endosteal surface, but rather that there is a preferential, geometrically specific deposition of bone.

Table Table 3.. Increased Endosteal Bone Formationin TSP2-Null Micea
Morphometric parameterMean (SD)Mean (SD)
  1. MAR, mineral apposition rate; MS, mineralizing surface.

  2. a Six-month-old males.

  3. *p < 0.05; n = 6.

Marrow area (mm2)0.80 (0.09)0.73 (0.16)
Endosteal perimeter (mm)3.38 (0.402)3.21 (0.354)
Endosteal MS (%)34.1 (8.54)53.6 (18.7)*
Endosteal MAR (μ/day)1.21 (0.24)1.30 (0.18)
Endosteal BFR (μ32 per day)0.41 (0.08)0.69 (0.21)*
Periosteal perimeter (mm)4.84 (0.164)4.78 (0.678)
Periosteal MS (%)58.12 (21.6)42.49 (22.03)
Periosteal MAR (μ/day)1.12 (0.42)1.07 (0.398)
Periosteal BFR (μ32 per day)0.72 (0.34)0.53 (0.32)

Mechanical testing showed no significant differences in strength, yield load, or failure load between the knockout and WT 4-month-old female mice (Table 4). Mechanical properties in bending are related directly to moment of inertia, which is a mathematical representation of the distribution of material from the neutral bending axis of the bone. In this calculation, the relative significance of material is weighted by the distance from the centroid; in other words, addition of material has a greater effect on moment of inertia if it is farther from the centroid of a section. Although there was a general decrease in the periosteal circumference, we found that the increased bone deposition on the endosteal surface is sufficient to maintain an equivalent moment of inertia (Table 4). Thus, the increase in endosteal bone formation, when coupled with a reduction in periosteal and endosteal perimeters, results in mechanical behavior that is relatively unchanged from that of controls.

Bone resorption in response to ovariectomy

Increases in bone formation do not necessarily preclude alterations in bone resorption.(32) Therefore, we conducted an experiment on 5-month-old female mice to determine whether TSP2-null mice were capable of a normal bone resorptive response when deprived of estrogen. The results show that mutant mice are as capable of bone resorption as are WT mice (Table 5). The percent changes in bone geometry and density, as determined by pQCT, in WT and TSP2-null ovariectomized mice essentially were the same.

Marrow-derived osteoprogenitor cells

Because the primary defect in TSP2-null mice is associated with specific alterations in endosteal bone, we hypothesized that there might be an alteration in the number of endosteal osteoblast precursor cells or in their synthetic activity. Endosteal osteoblast precursor cells arise from stromal cells that lie adjacent to the endosteal envelope in the marrow cavity.(37) These cells have been studied extensively and a direct relationship has been shown between the number of marrow stromal precursors and the number of colonies formed (termed CFU-F) when marrow suspensions are plated in vitro.(38) We determined both the total number of colonies and the number of AP-staining colonies on nine different groups of cells in duplicate. AP colonies represent MSCs that have begun to differentiate to preosteoblasts(39) Although there were large standard errors, which reflect the intrinsic variability of this in vitro analysis, a 2-fold increase was found in both the total number of CFU-F and the number of AP colonies (Fig. 4A). However, the ratio of AP to total colonies was not altered in TSP2-null animals. As well, the mean number of total harvested marrow cells was not statistically different (TSP2-null = 6.87 × 107 cells; WT = 6.41 × 107 cells) and mean body weight of the mice was the same (17.3 g for both TSP2-null and WT). Because increased colony numbers likely reflect an increased rate of proliferation in vivo, we evaluated the size of formed colonies as a preliminary indicator of proliferation rate(40) TSP2-null colonies were twice the size of WT colonies (Fig. 4A), a finding that suggests a greater rate of proliferation of the single-cell-derived clones.

Serum was obtained from mice at the time of harvest of MSC, and serum osteocalcin levels were measured. TSP2-null mice have a significant 35% increase (p < 0.05) in serum osteocalcin (data not shown), which is a sensitive indicator of the number of active osteoblasts.(41) The number of CFU-F from both WT and TSP2-null mice correlated significantly with serum osteocalcin levels (Fig. 4B), a finding that is consistent with the direct relationship between the latter parameter and the total number of osteoblasts.

Proliferation of MSCs in vitro

The increases in osteoblast precursors and in CFU-F size in TSP2-null mice suggested to us that there might be an increased proliferation of MSCs, both in vivo and in vitro. To further test whether in vitro proliferation was increased, serum-stimulated first-passage MSCs were labeled with 3H[thymidine]. Additionally, a fraction of first-passage cells was induced to become osteoblasts with BGP and Dex(42) and the proliferation of these cells also was evaluated. The results of these experiments showed a 2.5-fold increase in serum-stimulated DNA synthesis after serum starvation in TSP2-null MSCs (Fig. 5). However, MSC-derived osteoblasts from WT and TSP2-null mice showed similar levels of thymidine incorporation, a finding that suggests that the proliferation of mature osteoblasts is not different. Further analysis of MSC-derived osteoblasts also indicated that there were no differences in AP activity or in osteocalcin production per cell (data not shown). Thus, TSP2-null and WT endosteal-derived osteoblasts also appear to be equivalent synthetically.


Mice that lack the matricellular protein TSP2 have a unique alteration in long-bone modeling, which is characterized by an increased deposition of endosteal bone. Therefore, cortical thickness and density are increased and marrow area and periosteal and endosteal perimeters are reduced. At the same time, TSP2-null periosteal bone formation is decreased slightly. This reduction in periosteal bone formation presumably reflects a decreased requirement for weight-associated periosteal bone modeling, because increased endosteal bone deposition by TSP2-null mice results in bone strength equivalent to WT mice. Further analysis by μCT revealed that these alterations are site specific around the circumference of the femoral diaphysis. The increase in thickness and decrease in outer and inner fiber length are associated primarily with the medial and posterolateral aspects of the TSP2-null femoral diaphysis. It is likely that the increased deposition of bone at these specific sites is in part related to tendon and muscle strain (posterolateral deposition) and to modeling associated with weight bearing (medial deposition).

Figure Fig. 3..

Alterations in inner and outer fiber lengths and cortical thickness in circumferential intervals of the TSP2-null femoral diaphysis. Diaphyseal segment 1 (Fig. 1) was evaluated at 10° intervals around the central axis, yielding 36 interval measurements around the longitudinal axis. (A) Outer fiber lengths; (B) inner fiber lengths; (C) cortical thickness. *0.05 > p > 0.005; **p < 0.005; n = 5.

Table Table 4.. TSP2-Nulland WT Bones Are Mechanically Similar
ParameterMean (SD)Mean (SD)
  1. Iyy, moment of inertia; J, rotational moment of inertia; n = 5.

  2. a Measurements from the midsegment (segment 2, Fig. 1) of the diaphysis.

Stiffness (N/mm)189.5 (32.8)202.3 (19.4)
Yield Load (N)26.80 (5.08)24.52 (2.86)
Failure load (N)30.47 (4.71)28.22 (2.91)
Iyy (mm4)a0.17 (0.01)0.15 (0.03)
J (mm4)a0.42 (0.04)0.38 (0.07)
Table Table 5.. TSP-2 Null Mice Are Capableof Bone Resorption Inducedby Ovariectomy
 WT percent changeTSP2-null percent change
ParametersMean (SD)Mean (SD)
  1. There are no statistically significant differences in the percent change with ovariectomy (OVX) between WT and TSP2-null mice.

  2. *p < 0.05 between the OVX and sham-operated mice within the genotype (TSP2-null or WT); WT, n = 5; TSP2-null, n = 4.

Uterine weight− 43.2 (6.9)*− 45.8 (4.6)*
Cortical thickness− 12.0 (5.6)*− 7.5 (7.5)*
Cortical area− 8.9 (6.5)*− 5.9 (7.1)*
Periosteal diameter+ 0.1 (6.4)+ 0.4 (2.6)
Endosteal diameter+ 10.8 (15.2)+ 5.8 (6.9)
Cortical density− 2.5 (2.6)*− 3.2 (2.2)*
Total density− 10.0 (8.8)*− 9.2 (7.4)*
Figure Fig. 4..

MSCs are increased in number in TSP2-null mice and are correlated with osteocalcin levels. (A) Total marrow cells were harvested and plated at a density of 1 × 105 cells/cm2, and AP-positive and total colonies were evaluated; additionally, colony size was evaluated digitally using IQ. TSP2-null mice have a 2-fold increase in total and AP CFU-F and TSP2-null CFU-F are twice as large as WT. *p < 0.01; **p < 0.001; n = 9. (B) Serum osteocalcin levels are correlated with MSC numbers, as judged by CFU-F. Hatched squares = WT; open circles = TSP2-null (R = 0.667; p < 0.05; n = 12).

The increase in endosteal bone, observed in mice lacking TSP2, is associated with the presence of an increased number of marrow-derived osteoprogenitors. Interactions between osteoblast precursors and ECM components are critical for proper bone formation. Growth factors that are sequestered in the pericellular environment (43–45) and structural matrix macromolecules(46) both influence the differentiation and proliferation of MSCs. The proliferation of preosteoblasts in vivo is highly correlated with increased BFR and mineralizing surface.(47) Furthermore, the extent of mineralizing surface reflects the number of osteoblasts that are involved in bone formation.(48) We have found that TSP2-null mice have an increased mineralizing surface as well as an increased BFR and increased numbers of MSCs. It is highly probable that the increased bone formation is a consequence of the increased number of endosteal osteoblasts that arise from an enlarged pool of MSCs. Other instances of changes in MSC number in disordered bone remodeling have been reported. As examples, mice with age-associated osteopenia have a decreased number of MSCs,(39) whereas estrogen-depleted mice have increased numbers of MSCs.(32)

Growth factors produced by hematopoietic precursors(49) or of stromal cell origin increase both CFU-F formation and MSC proliferation. bFGF is especially mitogenic for MSCs in vitro,(45,50–52) and increases endosteal bone production in vivo.(53,54) PDGF also increases MSC CFU-F,(44,45) MSC proliferation,(55) and in vivo bone formation(56) The effects of TGF-βs are more complex. In vitro, TGF-β1 increases proliferation of rat MSCs(57) and TGF-β2 increases proliferation of human MSCs;(58) however, TGF-β1 also has been shown to decrease proliferation and increase differentiation of rat and pig MSCs.(50,51) MSC proliferation also is affected by systemic hormones. Vitamin D stimulates MSC proliferation,(59) whereas Dex has been shown to have species-specific effects, reducing rat MSC proliferation(51) but not affecting human MSC proliferation.(40) We also have found that the administration of Dex to mouse MSCs does not reduce 3H[thymidine] incorporation. These differing responses to Dex are not surprising because there are also species and system-specific differences in the differentiation response of osteoblasts or preosteoblasts to Dex.(60,61) In our mouse MSC system Dex appears to be incapable of decreasing thymidine incorporation into WT MSCs but can reduce the incorporation into TSP2-null MSCs to levels equivalent to those in WT cells.

Figure Fig. 5..

TSP2-null MSCs show increased DNA synthesis. The proliferation of first-passage MSCs was evaluated by measuring the incorporation of 3H[thymidine] after a period of serum stimulation. A fraction of the cells was differentiated to form osteoblasts on day 3 by the addition of BGP and Dex to culture medium. *Indicates value significantly different from all other values at p < 0.01; n = 6.

The effects of ECM proteins on MSC function have also been studied.(62) The culture of stromal cell lines on various collagen preparations influences the expression of the osteo-phenotype depending on the stromal cell line used and the collagen preparation, although an effect on cellular proliferation was not reported.(46) Tenascin-C is synthesized by MSCs(63) and may play a role in inducing osteoblast differentiation;(64) however, tenascin-C knockout mice have hematopoietic defects but no reported defect in MSCs.(65) We have results from immunocytochemistry and reverse-transcription polymerase chain reaction (RT-PCR) indicating that WT MSCs produce TSP2 in vitro (data not shown), but the reason for the increased proliferation of TSP2-null MSCs in vitro is not known and currently is under investigation.

TSP2, as a secreted matricellular protein, may influence MSCs in several different ways, but it is unlikely that decreased bone structural integrity is a factor in the response of TSP2-null MSCs. Compensation for structural weakness would be expected to occur at the periosteal surface of the bone rather than at the endosteal surface.(66,67) TSP2-null mice deposit extra bone at the endosteal surface and have a decrease in periosteal bone deposition. Furthermore, if extra bone were deposited to support a structurally weaker matrix, then it would be expected that the moment of inertia would increase correspondingly. However, we have found that the moment of inertia and mechanical strength are equivalent between TSP2-null and WT femurs. Additionally, electron microscope (EM) studies have shown that the collagen matrix of bone in TSP2-null mice is qualitatively normal (data not shown).

Thus, it appears likely that TSP2 influences the proliferation of MSCs in vitro and in vivo by modulating their biological interactions rather than by influencing the structural integrity of bone. For example, TSP2 could interact with cell-surface receptors on MSCs to regulate their proliferation. Alternatively, TSP2 could influence MSCs by inhibiting the vascularization of the bone marrow. Its absence would therefore result in increased blood supply to the marrow stroma and, possibly, lead to increased proliferation of MSCs. TSPs have been investigated extensively as negative modulators of blood vessel formation.(25) The exogenous administration of TSP2 decreases endothelial cell migration and proliferation(68) and inhibits the angiogenic response to bFGF in the rodent corneal assay.(69) As previously indicated, TSP2-null mice also have increased vessel density in a number of tissues.(4) However, if increased angiogenesis is the cause of increased proliferation of MSCs, it is unlikely that MSCs would continue to show a proliferative phenotype in vitro.

Another explanation for the growth characteristics of TSP2-null MSCs is that TSP2 functions indirectly in the bone marrow stroma and in culture by modulating the activity of growth factors. Other molecules that bind growth factors have been shown to regulate growth factor-mediated effects in bone. Alterations in systemic levels of insulin-like growth-factor-binding proteins increase bone formation in vivo in mice(70) and humans,(71) and the disruption of the biglycan gene, which encodes a proteoglycan present in bone that is believed to function in the regulation of TGF-β1, leads to osteoporosis(72) TSP1 binds and activates latent TGF-β1(21,22) and there is evidence that the phenotype of TSP1-null mice is, in part, related to decreased activation of TGF-β1.(9) TSP2 peptides inhibit the activation of TGF-β1 by TSP1;(73) thus the absence of TSP2 may increase levels of active TGF-β1. TSP1 also binds PDGF(24) and bFGF.(23,74) Although the PDGF- and bFGF-binding domains of neither TSP1 nor TSP2 have been identified, the amino acid sequence similarity between the two proteins raises the possibility that TSP2 could sequester bFGF or PDGF and inhibit interactions with their receptors.

Although the mechanism by which TSP2 influences MSCs remains unclear, its absence in mice results in a biologically driven, unique alteration in bone modeling. We have shown that the lack of TSP2 leads to increased cortical density and thickness with an associated decrease in marrow volume and a decrease in endosteal and periosteal diameters. Our findings suggest that enhanced endosteal bone production in the absence of TSP2 is an alternative modeling mechanism that can be used to achieve bone strength equivalent to that in normal mice. A reduction in the in vivo activity or synthesis of TSP2 might represent a means to increase endosteal bone mass and enhance the mechanical strength of bone.


We thank Dr. Mark McKee (McGill University) for EM analyses of bone. We also thank Debbie Puerner, Kirsti Scheer, and Martha Strachan (Skeletech, Inc.) for their assistance with bone histology and pQCT. This work was supported by NIH grants HL 18645 (P.B.), AR 11248 (P.B.), and AR 34399 (S.A.G.) and by NIH Training grants DE 07063 (K.D.H.) and AG 00114 (E.A.S.).