The Secreted Protein Thrombospondin 2 Is an Autocrine Inhibitor of Marrow Stromal Cell Proliferation


  • Kurt D. Hankenson,

    1. Department of Biochemistry, University of Washington, Seattle, Washington, USA
    2. Present address for Dr. Hankenson: Orthopaedic Research Labs, University of Michigan, G161-400 N. Ingalls Bldg., Ann Arbor, Michigan, USA
    Search for more papers by this author
  • Paul Bornstein M.D.

    Corresponding author
    1. Department of Biochemistry, University of Washington, Seattle, Washington, USA
    2. Department of Medicine, University of Washington, Seattle, Washington, USA
    • Department of Biochemistry, University of Washington, Box 357350, Seattle, WA, USA 98195-7350
    Search for more papers by this author

  • Presented in part in preliminary form as a short communication in Trans Orthop Res Soc 2001, p. 620.

  • The authors have no conflict of interest.


Marrow stromal cells (MSCs) are obtained in increased number from mice in which the thrombospondin 2 (TSP2) gene is disrupted, and these cells show increased DNA synthesis in vitro. To examine more closely the role of TSP2 in the physiology and osteogenic differentiation of MSCs, an in-depth characterization of TSP2-null MSCs was conducted. Determination of TSP2 protein content by Western analysis and RNA levels by reverse-transcription polymerase chain reaction (RT-PCR) indicated that MSCs are the primary source of TSP2 in the marrow and secrete abundant TSP2 into culture medium. Morphologically, the TSP2-null and wild-type (WT) cell populations were similar and by flow cytometry contained equivalent numbers of CD44+, Mac1+, intercellular adhesion molecule-1 (ICAM-1+), and ScaI+ cells. TSP2-null cells showed delayed mineralization associated with an increased rate of proliferation. Consistent with this finding, there was a decrease in expression of collagen and osteocalcin RNA by TSP2-null MSCs on day 7 and increased osteopontin expression on day 7 and day 14. In add-back experiments, recombinant TSP2 produced a dose-dependent decrease in proliferation. This reduction was associated with an accumulation of TSP2-treated cells in the G1 phase of the cell cycle and did not result from an increase in apoptosis. When TSP2 treatment was terminated, the cell population reentered the S phase. We conclude that the increased endosteal bone formation observed in TSP2-null mice results primarily from the failure of TSP2 to regulate locally MSC cell cycle progression.


SUBSTRATUM-ADHERENT CELLS obtained from whole marrow, referred to as marrow stromal cells (MSCs), mesenchymal stem cells, or marrow fibroblasts, are osteoblast progenitors that are sequestered within the endosteal envelope.(1) When MSCs are cultured in vitro, they express osteogenic proteins(2) and the osteogenic transcription factor cbfa1(3) and form a mineralized matrix.(4) When ceramic carriers containing purified MSCs are transplanted subcutaneously in mice, MSCs form bone with a marrow cavity without the addition of exogenous osteogenic factors.(5) On transplantation to irradiated mice, MSCs differentiate to become osteoblasts and osteocytes, which can be identified in the lacunae of cortical bone.(6)

The quantity of MSCs in the marrow cavity, as determined by a colony-forming unit fibroblastic (CFU-F) assay,(7) is correlated with the extent of bone formation. Growth factors such as transforming growth factor β1 (TGF-β1)(8) and systemic factors such as parathyroid hormone (PTH)(9) when administered in vivo increase CFU-F and bone formation. Estrogen depletion due to ovariectomy increases CFU-F,(10) and bone loss that occurs as a result of decreased weight bearing(11) or aging(12) is associated with decreased CFU-F. Various genetically engineered mice, such as the biglycan,(13) peroxisome proliferator-activated receptor-α (PPARα),(14) and fibroblast growth factor (FGF) 2 knockout mice,(15) show differences in bone formation that are associated with changes in CFU-F number. Mice with a targeted disruption of the thrombospondin-2 (TSP2) gene also have an increased formation of endosteal bone that is correlated with an increase in the number of MSCs.(16) Furthermore, TSP2-null MSCs, when cultured in vitro, show increased DNA synthesis as determined by tritiated thymidine incorporation.(16)

Proliferation of MSCs in vitro is sensitive to a variety of extracellular molecules including systemic hormones such as vitamin D(17) and dexamethasone(18) and growth factors(19) such as basic FGF (bFGF),(20) platelet-derived growth factor (PDGF),(21) and the bone morphogenetic protein secreted protein (BMP) superfamily members.(22,23) The effect of extracellular matrix (ECM) proteins on MSC proliferation has not been studied as extensively.(24) Culture of stromal cell lines on various collagen preparations influences the expression of the osteophenotype, although changes in cellular proliferation were not reported.(25) Tenascins, including tenascin C, are immunolocalized to MSCs in vivo(26); however, tenascin-C knockout mice have hematopoietic defects with no reported defect in MSCs.(27)

TSP2 is a 450-kDa trimeric extracellular protein that is expressed in osteogenic tissue and in the MC3T3-E1 osteoblast cell line.(28) The protein is one of five members of the TSP family, which are all expressed in bone or cartilage.(29) TSP2 and TSP1 are structurally similar, but there is considerable divergence from the other three family members.(29) TSP1 and TSP2 act as extracellular modulators of cell function, rather than serving a primary structural role, and in this capacity the two proteins have been termed matricellular proteins.(30) Interestingly, a number of other matricellular proteins are expressed also in bone, including secreted protein, acidic and rich in cysteine (SPARC) (osteonectin), tenascin C, and osteopontin. As modulators of cell function, these proteins bind to exogenous growth factors, matrix metalloproteinases, and structural matrix proteins; however, they also interact directly with cells via a variety of cell-surface receptors, including integrins. Although there are no published reports of TSPs playing a role in regulating MSC function, TSP1 and to a lesser extent TSP2, have been studied extensively as inhibitors of endothelial cell (EC) growth and migration.(31)

Here, we show that MSCs produce TSP2 and that TSP2-null MSCs show an increased rate of proliferation but delayed osteogenesis relative to wild-type (WT) cells. When TSP2 is added to MSCs in culture, it inhibits proliferation in a dose-dependent manner by decreasing the percentage of cells that are in S phase, without affecting apoptosis. Thus, we conclude that TSP2 is an autocrine inhibitor of MSC cell cycle progression and that its antiproliferative activity promotes earlier in vitro mineralization.



The generation of TSP2-null mice has been reported previously.(32) TSP2-null and WT colonies were housed separately under a constant 12-h light/dark cycle in a specific pathogen-free facility. The TSP2-null and congenic WT mice exist in both mixed C57BL/6J × 129/SvEms-+Ter and pure 129 backgrounds. All procedures have been approved by the Institutional Animal Care and Use Committee at the University of Washington.


Full-length TSP2, amino-terminal TSP2 (nTSP2; containing amino acids 1-296) and the procollagen domain of type I collagen (Col1) were produced in a baculovirus expression system. TSP2 and nTSP2 were purified on a heparin-Sepharose column, and Col1 was purified using an S-100 gel filtration column. Columns and the enhanced chemifluorescence kit were obtained from Amersham Pharmacia Biotech (Piscataway, NJ, USA). Purified proteins were assayed for endotoxin using a Limulus amebocyte lysate assay (Associates of Cape Cod, Falmouth, MA, USA) and used only when endotoxin levels were below the levels found in bovine serum. Antibodies for flow cytometry, Fc block, and ApoDirect were from Pharmingen (San Diego, CA, USA). Sybr Green I, Cyquant assay, and propidium iodide (PI) were obtained from Molecular Probes (Eugene, OR, USA). The Qiamp DNA purification kit, Omniscript reverse transcriptase, and Hot-Star Taq DNA polymerase were from Qiagen (Valencia, CA, USA). All other reagents were obtained from Sigma (St. Louis, MO, USA).

MSC culture

Cells were harvested from the bone marrow of adult mice according to previously published procedures.(16) Primary cells were cultured in MSC medium (α-minimum Eagle's medium, 10% fetal calf serum, 25 μg/ml of sodium ascorbate, 100 IU/ml of penicillin, 100 μg/ml of streptomycin, and 10 μM of amphotericin-B) for 12-14 days. The medium was changed on day 4 and then every third day. MSCs were harvested using 0.25% trypsin/10 mM of EDTA and plated at cell densities of 1 × 104 cells/cm2. For mineralization experiments, treatment with mineralization medium (MSC medium + 10 mM of β-glycerophosphate) began on day 4.

Western blot

Media were harvested from 8-day cultures of first-passage MSCs, concentrated 1:3 using Centricon-50 filters (Millipore, Bedford, MA, USA) and frozen at −20°C. The cell layers were lysed and DNA was purified using the Qiamp kit according to the manufacturer's instructions. DNA content was measured by absorbance at A260. A linear concentration range of purified TSP2 (rTSP2) and equal volumes of medium from three independent lines of WT cells and a TSP2-null cell line were loaded onto 7% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels. Protein was electrophoretically transferred to nitrocellulose. The blot was blocked with phosphate-buffered saline (PBS)/0.1% Tween/5% powdered milk and incubated with polyclonal rabbit anti-TSP2 antibody. A goat anti-rabbit alkaline phosphatase (ALP)-conjugated secondary antibody was used for detection of fluorescence using enhanced chemi-fluorescence (ECF) reagent. Immunoreactive TSP2 bands were visualized using a Storm Imager (Molecular Dynamics, Sunnyvale, CA, USA) and band intensities were quantified using ImageQuant software (Molecular Dynamics). A standard curve was generated based on the signal intensity of known concentrations of purified TSP2.

TSP2 reverse-transcription-polymerase chain reaction

Guanidine hydrochloride was used to extract RNA from three different populations of cells: (1) first-passage MSC (day 7, postplating), (2) whole marrow flushed from tibias and femurs, and (3) marrow cells that do not adhere to a plastic substratum 12 h after plating. Two micrograms of RNA was reverse-transcribed (RT) in 30-μl reactions using Omniscript and 5 μl of complementary DNA (cDNA) was used in 100-μl polymerase chain reactions (PCRs) for 20 cycles or 30 cycles for 1 minute at 94°C, 1 minute at 55°C, and 2 minutes at 72°C with a 15-minute hot start at 95°C. The primers for TSP2 were F-CTG GTG ACC ACG TCA AGG ACA CTT CAT and R-ATG CAC CTT TGG CCA CGT ACA TCC TGC, and the primers for the ribosomal protein gene S6, were F-AAG CTC CGC ACC TTC TAT GAG A and R-TGA CTG GAC TCA GAC TTA GAA GTA GC. Amplified DNA was run on 2% agarose minigels and DNA was stained with Sybr Green I and visualized with a Storm fluorescence imager (Molecular Dynamics, Sunnyvale, CA, USA).


The adherent cells from whole marrow were grown until day 14 and harvested by trypsin digestion in the presence of EDTA. Cells were counted and a million cells for each antibody were pelleted. Cells were blocked with Fc block in PBS/2% bovine serum albumin (BSA) for 30 minutes and then exposed to fluorescein isothiocyanate (FITC; ScaI, Mac1, ICAM-1, and syndecan)- or phycoerythrin, PECAM, CD44, B220, and Thy1.2)-conjugated primary antibodies. Stained cells were examined using flow cytometry (Becton-Dickson, Franklin Lakes, NJ, USA) and gated based on size and granularity. Fluorescence of gated cells was visualized as a histogram of fluorescence intensity versus event number. A gate was set based on comparison with isotype-matched negative controls to determine the percentage of the cells expressing a given antigen.

In vitro mineralization

Mineralization of β-glycerophosphate-treated cells was evaluated by staining with 1% alizarin red after fixation in 75% ethanol. Alizarin-stained plates were imaged using a flatbed scanner (Agfa-Gevaert N.V., Mortsel, Belgium).

Semiquantitative RT-PCR

RNA from MSCs, cultured for varying times (7, 14, and 21 days), was harvested by guanidine hydrochloride extraction. Primer sets were based on published reports for glyceraldehyde-3-phosphate dehydrogenase (GAPDH), cbfa1, ALP,(33) and osteocalcin.(34) Col1A1 (F-CCA CGC ATG AGC CGA AGC TAA CCC C and R-CCG GGC TTG CCA GCT TCC CCA TCA TC), TSP1 (F-TGG TGT CAG TGG AGG AGG CT and R-TGA CCA CTT TTC GGA TGC TGT), SPARC (F-CCC CTC AGC AGA CTG AAG TT and R-GGC AGG AAG AGT CGA AGG TC), and osteopontin (F-CTG GCA GCT CAG AGG AGA AG and R-AGG TCC TCA TCT GTG GCA TC) primers were designed based on the published cDNA sequences. First-strand synthesis was performed with Omniscript using 1 μg of RNA/30 μl of reaction. First-strand cDNA (0.5 μl) was used in 20 μl PCR reactions with 300 pg of primer/reaction using HotStar Taq DNA polymerase. The correct cycle number for each primer set that resulted in signal linearity was determined by examining first-strand synthesized products over a 100-fold concentration range at varying cycle numbers. Eight microliters of PCR product was loaded onto 2% agarose gels, run at 100 V, and stained with Sybr Green I. Band intensity was evaluated on the Storm fluorescence imager, and each sample was standardized by determining the signal ratio relative to both GAPDH and S6. For each transcript, a 100% expression level was arbitrarily assigned to the day 7 WT MSC values. TSP2-null values at days 7, 14, and 21, and WT values at day 14 and day 21 were compared with the day 7 WT value to generate relative levels of expression.

Cell proliferation

Cells were either counted directly using a hemocytometer or a fluorescence assay was used to determine cell number. For direct cell counts, cells were plated at 2 × 105 cells/well in 6-well plates and harvested by trypsinization at indicated time points. For Cyquant determination of proliferation, cells were plated at 1 × 105 cells/well in 12-well plates. Plates were harvested on day 2, a time when very little proliferation had occurred and when TSP2-null and WT cells did not show significant variability in cell number. Parallel culture plates, either control or treated, were then harvested on day 4 or day 6. Harvested plates were rinsed with PBS, patted dry, and then frozen at −70°C pending evaluation. Frozen plates were thawed and cells were lysed using 1 ml of Cyquant lysis buffer. A 50-μl aliquot, in duplicate, was transferred to wells of a 96-well plate and 50 μl of 2× Cyquant reagent was added to each well. Fluorescence intensity was measured using the Storm imager with ImageQuant software, and the increase in signal relative to control cells was determined.

Cell cycle analysis and apoptosis

MSCs (5 × 105 cells/plate) were plated onto 60-mm tissue culture plates. Seventy-two hours after plating, media were changed and cells were treated with 50 nM of TSP2. Cells were harvested at 48 h. Pelleted cells were resuspended in 75% ethanol, counted directly using a hemocytometer, and stored at 4°C until analyzed. For PI staining, pelleted cells were resuspended in 1 ml of PI solution + 100 U RNAse A. Cells were stained for more than 1 h at 4°C and then analyzed by flow cytometry to obtain a histogram of DNA content relative to cell number. The percentage of cells in sub-G1, G1, S, or G2/M was determined by analyzing the data using ModFitLT software (Verity Software, Topsham, ME, USA). To determine apoptosis, harvested cells were fixed in 1% paraformaldehyde and then stored in 75% ethanol at 4°C. Fragmented DNA within cells was end-labeled with deoxyuridine triphosphate (dUTP)-FITC according to the manufacturer's instructions using the ApoDirect kit. FITC+ and PI-stained cells were visualized and compared with apoptosis-positive and -negative controls (provided by the manufacturer) using two-color sorting.

Statistical analysis

Statistically significant differences at p < 0.05 were determined using one-way analysis of variance (ANOVA) procedures (Excel; Microsoft, Redmond, WA, USA).


TSP2 production by MSCs

TSP2 was secreted by MSCs, as evidenced by its presence on Western blot (Fig. 1). TSP2-specific bands were present in the recombinant controls and in WT conditioned media, but there was no corresponding band in the TSP2-null sample. With an increase in cell number, there was an increase in TSP2 concentration in the media, but the amount of TSP2 produced on a per cell basis remained equivalent, with a mean value of 4.2 μg/1 × 105 cells. The diffuse, less-intense bands observed in the TSP2-null lane are caused by nonspecific binding of the anti-TSP2 polyclonal antibody with serum proteins; these bands, located above and below the TSP2 immunoreactive bands, also are observed in the WT lanes. Interestingly, the recombinant TSP2 produced by insect cells migrated slightly faster in SDS-PAGE than murine-derived TSP2, possibly because of alterations in glycosylation.

Figure FIG. 1..

TSP2 is secreted by MSCs in culture. Media were harvested from day 8 cultures of three different WT and one TSP2-null cell preparations at varying densities (WT-1, 300,000; WT-2, 600,000; WT-3, 1,200,000; null, 1,000,000 cells/plate). Media were concentrated 3:1 using a Centricon filter with 50 kDa MW cut-off. Standard curves were made with purified recombinant TSP2 protein. Western blots were probed with rabbit anti-amino-terminal TSP2 antibody followed by goat anti-rabbit ALP-conjugated secondary antibody. Fluorescence of bands was compared with a standard curve (R2 = 0.967). Corrected values, WT-1 = 1.9 μg/ml; WT-2 = 3.9 μg/ml; WT-3 = 6.9 μg/ml. The result shown is representative of three separate experiments.

TSP2 in the marrow could be derived from a variety of cell types in addition to MSCs. To examine whether TSP2 is produced by nonadherent hematopoietic cells, RT-PCR was used. Although a robust signal for TSP2 was evident in the MSCs at 20 cycles, a cycle number that was determined to result in linear Sybr Green I staining over a 100-fold cDNA concentration range, no comparable signal was observed with whole marrow RNA or with RNA obtained from the nonadherent cell (NAC) population (Fig. 2A). To determine whether TSP2 transcripts could be present in whole marrow or in NACs at levels that are undetectable in the linear range of amplification, PCR was extended for another 10 cycles (30 total). Under these nonlinear conditions, a TSP2-amplified band was present; however, both bands were less intense than the MSC signal (Fig. 2B). This result suggests that MSCs are the primary source of TSP2 in the marrow environment; however, TSP2 also may be produced at lower levels by a subset of nonadherent hematopoietic progenitor cells. Conversely, the TSP2 signal visualized at a higher cycle number could be the result of RNA contributed by MSCs contained within whole marrow or because of MSCs that remained unattached to the substratum 12 h postplating.

Figure FIG. 2..

TSP2 is primarily derived from MSCs in the marrow environment. RNA from first-passage MSCs on day 7, whole marrow, or NACs was reverse-transcribed and amplified by PCR for (A) 20 cycles or (B) 30 cycles. Samples were loaded on 2% agarose gels and DNA was visualized using Sybr Green I on a fluorescence imager. Three samples were analyzed.

Increased proliferation of MSCs in the absence of TSP2

Although we previously showed that first-passage MSCs have an increased incorporation of tritiated thymidine,(16) rates of proliferation were not examined. TSP2-null and WT cells, plated at an equivalent density, showed a similar increase in proliferation until day 4 or day 5. After that time, TSP2-null MSCs had a 2.7-fold greater rate of proliferation than WT cells (Fig. 3). Cell proliferation assays were routinely carried out from day 2 to day 11 or day 12; however, at a later time (day 21) TSP2-null cells were still increased in number (data not shown). We observed that WT cells become more spread than the TSP2-null cells over time in culture, suggesting that the WT MSCs might become quiescent. This was not caused by a limited capacity to divide, because when WT cells were exposed to a higher level of serum (20%), they were able to divide further to yield an increased number of cells (data not shown).

Figure FIG. 3..

MSCs show a time-dependent increase in cell number. TSP2-null and WT cells were plated at 2 × 104 cells/cm2 in wells of 6-well plates and harvested by trypsinization for counting. Medium was changed every 3 days. The rate of proliferation for TSP2-null cells from day 3 to 11 is 82,500 cells/day, and the WT cells increased at a rate of 31,150 cells/day (n = 4, *p < 0.05).

Phenotypic characterization of TSP2-null and WT MSCs

Murine MSCs express a variety of phenotypic markers, suggesting that the cell population is not completely homogeneous. In addition, first-passage MSC cultures contain contaminating monocytes.(35,36) Thus, differences in growth between WT and TSP2-null cells could be attributed to differences in the subpopulations of cells that are present. Based on histological examination, WT and TSP2-null cells were morphologically similar and had a similar number of fibroblastoid cells that were ALP+ (Fig. 4). The fibroblastoid cells generally were present in sheets of tightly packed cells, whereas cells presumed to be monocytes were distributed randomly.

Figure FIG. 4..

WT and TSP2-null MSC are ALP+ and morphologically similar. Cells were plated at 2 × 104 cells/cm2 in 35-mm plates. The cells were harvested on day 3 and stained using ALP histochemistry (ALP+ cells are stained purple) or a modified Giemsa stain. Asterisks denote the cytoplasm of MSCs. Arrows indicate nonfibroblastoid monocytes.

To obtain a more definitive measure of cellular phenotypes present in WT and TSP2-null MSC cultures, immunophenotyping by flow cytometry was used. There were two cell populations that could be separated based on size and granularity, a small agranular (SA) and a large granular (LG; Fig. 5A). These two populations had different background staining intensities, which necessitated a separate examination of their antigen expression. In both WT and TSP2-null samples, the SA cells represented the majority population (94.6% and 92.2%, respectively). The gated populations were examined for cell-surface markers reported to be present on MSCs (CD44, ScaI, ICAM-1, and Thy1.2) or non-MSC markers that are expressed on other marrow-derived cells (Mac1, PECAM, syndecan, or B220).(37,38) Although MSC cultures were positive for Mac1 (CD11b, macrophage marker), the hyaluronate receptor CD44, ScaI (expressed on lymphohematopoietic cells as well as on MSCs), and ICAM-1 (expressed on lymphocytes, vascular endothelium, dendritic cells, and osteoblast lineage cells; Fig. 5B), there were no remarkable differences in expression between WT and TSP2-null (Table 1). There was no significant expression of Thy1.2 (a thymocyte marker reputedly expressed on human MSCs), PECAM (CD31 on ECs), syndecan (high levels on epithelial cells), or B220 (B-cell marker) in either the SA or LG populations. We conclude that WT and TSP2-null cultures are phenotypically equivalent and are composed primarily of MSCs, with contaminating monocytes, but not B cells, T cells, epithelial cells, or ECs.

Table Table 1.. Expression of Cell-Surface Antigens by WT and TSP2-Null MSCa
original image
Figure FIG. 5..

WT and TSP2-null MSC cultures are phenotypically similar. Primary cells were grown to day 14 and harvested by trypsinization. Cells were labeled with directly conjugated antibodies (FITC and PE) and examined using flow cytometry. (A) Two discrete cell populations could be identified based on size and granularity, an LG population (LA) and a more numerous population of SA cells. (B) Representative histograms showing the expression of Mac1 in the SA cells, ScaI in the LA cells, CD44 in the SA cells, and ICAM-1 in the SA cells. The isotype control is represented by a line, and the positive cells are represented by the filled-in curve. The percentage of positive cells was determined by calculating the percentage of cells that do not overlap with the isotype control. Flow cytometry was repeated on three different occasions with primary MSCs (n = 2 per experiment).

Osteogenic potential of MSCs in the absence of TSP2

Although our previous results suggest that an increase in bone formation develops in TSP2-null mice because of an increase in the number of osteoblast precursors,(16) it also is possible that TSP2-null osteoblasts have an enhanced osteogenic capacity. To assess osteogenesis, cells were treated with β-glycerophosphate on day 4 of culture and plates were stained with alizarin red on days 7, 14, 21, and 35. There was no mineralization present on day 7 in either cell type. On day 14, WT cells were positive for calcium deposition, but TSP2-null cells had little detectable mineralization. However, by day 21 the TSP2-null wells had formed mineral that appeared equivalent to that of WT (Fig. 6). By day 35, >95% of the surface area of both plates was stained with alizarin red. Cellular ALP levels did not reflect the delayed osteogenesis in TSP2-null plates because there were no differences in ALP activity between WT and TSP2-null MSC on days 7, 14, and 21 (data not shown).

Figure FIG. 6..

TSP2-null MSCs show a delay in mineralization. TSP2-null and WT cells were plated at 2 × 104 cells/cm2 in wells of 6-well plates, two of which are shown for each time point. Wells were treated with MSC medium containing 10 mM of β-glycerophosphate and medium was then changed every third or fourth day. Plates were examined on days 7 (no positive staining, not shown), 14, 21, and 35 by fixation in 75% ethanol and staining with alizarin red. Results shown are representative of those obtained in three independent experiments.

Mineralization in culture occurs in response to a specific temporal profile of gene transcription.(2,4,39) To determine whether the delay in mineralization of TSP2-null MSC was correlated with alterations in gene expression, transcription of eight genes expressed in osteoblasts was studied using semiquantitative RT-PCR at three different times (7-, 14-, and 21-day postplating). Both WT and TSP2-null cells expressed all of the genes of interest and showed significant, time-dependent variations in collagen, osteocalcin, SPARC, and TSP1 expression (Figs. 7A and 7B). Although there was a trend for the TSP2-null cells to maintain a higher level of expression for all of the osteogenic protein genes on day 21, the levels were only significantly different for collagen and osteocalcin, which were both increased 3-fold over WT (Fig. 7A). Interestingly, on day 7 the reverse was true; the levels for collagen and osteocalcin were significantly lower for the TSP2-null cells in comparison with the WT cells, 50% and 33% less, respectively. In contrast, the expression levels for the matricellular protein gene osteopontin were increased significantly on day 7 (3-fold) and day 14 (2-fold) for TSP2-null MSC in comparison with WT cells (Fig. 7B). The significance of these results will be addressed in the Discussion section.

Figure FIG. 7..

WT and TSP2-null MSCs show different patterns of RNA expression. RNA from WT and TSP2-null MSC harvested on 7, 14, and 21 days was reverse-transcribed. cDNA was amplified with primer sets for (A) osteogenic protein genes ALP (22 cycles), cbfa1 (28 cycles), Col1a1 (22 cycles), and osteocalcin (OC; 24 cycles) or (B) matricellular protein genes osteopontin (OPN; 20 cycles), SPARC (24 cycles), TSP1 (20 cycles), and TSP2 (22 cycles). For each RNA sample (n = 3; PCR reactions repeated in duplicate), the intensity of fluorescence was compared in a direct ratio to two “housekeeping” genes, S6 (26 cycles), and GAPDH (24 cycles). A relative percentage was determined by comparing values to the day 7 WT samples, which were arbitrarily set at a value of 100%. There were no significant differences in S6 or GAPDH expression between TSP2-null and WT cells at any time point (not shown; *p < 0.05).

Neither TSP1 nor SPARC, two matricellular proteins that also have been shown to decrease cell growth,(40,41) were increased in a compensatory manner in TSP2-null MSCs (Fig. 7B). Expression patterns for these two genes were similar between TSP2-null and WT cells, increasing several-fold from day 7 to 14 and then returning to day 7 levels on day 21. TSP2 levels, on the other hand, decreased gradually from day 7 to 21, without any increase on day 14 (Fig. 7B).

The effect of exogenous TSP2 on MSC proliferation

Because MSCs secrete TSP2, we hypothesized that TSP2 acts in an autocrine manner to decrease MSC proliferation. The addition of recombinant, insect cell-produced TSP2 to the culture medium decreased proliferation of both WT and TSP2-null cells (Fig. 8A). However, neither nTSP2 nor Col1, two protein fragments also produced in a baculovirus expression system by insect cells, decreased MSC proliferation. A more detailed analysis of TSP2-null MSCs showed that they have a dose-dependent, log-linear (R2 = 0.97) decrease in proliferation when exposed to recombinant TSP2 (Fig. 8B). The average EC50, determined in three separate experiments, was 39.2 nM of TSP2 monomer (∼3.0 μg/ml), a level similar to that which can be achieved in culture medium. Similar concentrations of nTSP2 did not decrease proliferation. A dose-response curve was determined also for ST2 cells, a murine MSC line,(42) and for immortalized TSP2-null MSCs, a stromal cell population devoid of monocytes. Immortalized TSP2-null MSCs were obtained from Immortomouse/TSP2-null double transgenic mice that were generated in our laboratory.(43) Both cell types also were sensitive to TSP2-mediated inhibition (data not shown).

Figure FIG. 8..

Exogenous TSP2 treatment results in a concentration-dependent decrease in proliferation of TSP2-null MSCs. Cells were plated in quadruplicate in 12-well plates at 1.0 × 105 cells/well and a set of cells harvested and frozen at −70°C on day 4. (A) Parallel wells were treated with 50 nM of TSP2, nTSP2, or Col1 and wells were harvested on day 8. Relative changes in cell content were determined using the Cyquant proliferation assay. Results are expressed as the percent increase relative to WT control (*p < 0.05 when untreated controls and experimental groups were compared for each genotype). The WT and TSP2-null controls also were significantly different (p < 0.05). (B) TSP2-null cells were treated with increasing concentrations of TSP2 or nTSP2 and harvested on day 8. Proliferation, relative to a buffer control, was determined and a dose-response curve was generated. Both experiments were repeated three separate times.

TSP2 might decrease cell growth by limiting the percentage of cycling cells or by inducing cell death. To examine the dynamics of the cell cycle, TSP2-null MSCs were cultured with TSP2 for 48 h and cells were examined for DNA content by flow cytometry. Treatment with TSP2 decreased cell number by 50% (not shown), and 50% fewer cells were in the S phase than in nontreated cells (Table 2); instead, TSP2-treated cells accumulated in G1. When TSP2 was removed from the cells for 48 h, the percentage of S phase cells was restored (Table 2). This finding suggests that TSP2 regulates the G1 to S phase transition, but that it does not affect the S to G2/M or G2/M to G1 transitions, because there is not an accumulation of cells at either of those transition points. In both treated and untreated populations, only a small percentage of cells were in the sub-G1 fraction, 3.87% and 2.81% of total cells, respectively, suggesting that very little apoptosis was occurring. To further determine whether TSP2 induced MSC apoptosis, terminal transferase-mediated dNTP nick-end labeling (TUNEL) staining was used to examine cells for DNA damage. Neither control cells nor TSP2-treated cells showed a significant level of TUNEL staining. Both had single peaks with mean fluorescence values of 4.24 and 5.69, respectively. In comparison, apoptosis-positive Jukart cells (provided by the manufacturer) showed a dual peak of nonapoptotic (78.4% of total cells) and apoptotic (21.6%) cells with mean fluorescence values of 0.76 and 210.2, respectively (data not shown).

Table Table 2.. TSP2 Decreases the Percentage of TSP2-Null MSCs That Are in S Phasea
original image


We have established that TSP2 is an autocrine factor that regulates the proliferation of marrow-derived osteoprogenitors, and thus provides a mechanistic link between increased bone formation and increased CFU-F observed in TSP2-null mice.(16) A similar antiproliferative effect of TSP2 has been described for ECs(44) and numerous studies have established the antiproliferative effect of TSP1 on non-MSC cell types.(40) Despite its obvious role in limiting proliferation, TSP2 appears to have no deleterious effect on in vitro osteogenesis. TSP2-null MSCs did not show an increased capacity for in vitro mineralization; in fact, WT MSCs, which produced abundant TSP2, mineralized earlier than TSP2-null MSCs. We hypothesize that the enhanced endosteal bone formation observed in TSP2-null mice is caused by primarily an increase in MSC proliferation that results in an increase in the number of endosteal osteoblasts and not by an increase in cellular osteogenic potential.

Deposition of an extracellular matrix rich in type I collagen and osteocalcin is required for in vitro mineralization,(2,4,39) and osteopontin has been shown to have a negative effect on mineral accumulation in culture.(45) Thus, calcium deposition by TSP2-null cells could be delayed because of the alteration in the normal pattern of collagen, osteocalcin, and osteopontin gene expression (Fig. 7). This aberrant gene expression is likely to result from the increased proliferation rate of TSP2-null cells, because numerous studies have shown that the ability of osteoblast-lineage cells to generate an ECM capable of mineralization is correlated inversely with proliferation.(39) The increased expression of collagen and osteocalcin by TSP2-null MSC on day 21 may compensate for the reduced expression that occurred earlier in culture.

We have tested the hypothesis that TSP2 enhances mineralization through its inhibition of MSC proliferation by treating both ST2 cells and TSP2-null immortalized MSCs with exogenous TSP2 during the process of mineralization. TSP2 inhibited proliferation of both cell lines in a dose-dependent manner (see Results section) and the addition of TSP2 resulted in earlier mineralization in both MSC lines (K. D. Hankenson and P. Bornstein, unpublished results, 2001). Future studies will be focused on determining whether accelerated mineralization in the presence of TSP2 can be uncoupled from proliferation inhibition.

Recent studies that have explored the mechanism of growth regulation by the TSPs have focused on the interaction of TSP1 with the cell surface receptor CD36.(46,47) The binding of TSP1 to CD36 activates a caspase-dependent apoptosis pathway in EC.(39) However, in MSCs we have found that apoptosis is not induced by TSP2 treatment; rather, TSP2-treated MSCs appear to be arrested at the G1/S boundary. TSP2 could still be acting through CD36, but via a nonapoptotic pathway, to decrease the number of cells entering the S phase. However, in a preliminary study, we have found that CD36-null MSCs are as sensitive to inhibition by TSP2 as are WT cells (K. D. Hankenson, M. Febbraio, R. Silverstein, and P. Bornstein, unpublished results, 2001). Thus, TSP2 appears to regulate the growth of MSCs by a mechanism different from that shown for regulation of EC growth by TSP1. These findings may reflect a physiological distinction between EC and MSC or a mechanistic difference between the actions of TSP1 and TSP2. Recent results from our laboratory do not answer this question conclusively because TSP2 both induces apoptosis and inhibits cell cycle progression in human microvascular EC (L. Armstrong, B. Bjorkblom, K. D. Hankenson, A. Siadek, and P. Bornstein, unpublished data, 2001).

Several β1- and β3-integrins and integrin-associated protein (CD47) also mediate TSP-cell interactions.(40) MSCs express multiple integrins(48) and the engagement of integrins has been shown to affect cell cycle progression.(49) However, the type I repeats that serve as the primary growth-inhibitory domain of TSP1(50) and TSP2 (K. D. Hankenson and P. Bornstein, unpublished results, 2001), do not contain an RGD sequence capable of binding ECM proteins to integrins. There is an additional integrin-binding domain located within the amino-terminus of TSP1 that binds α3β1, and the corresponding sequence in TSP2 is similar but not identical.(51) The interaction of TSP1 with α3β1 inhibits EC growth,(51) but we have shown that the amino-terminal domain of TSP2 (nTSP2) does not inhibit proliferation of MSCs (Fig. 8). Therefore, it is possible that TSP2 does not bind with high affinity to α3β1. The interaction of TSP1 with CD47 is responsible for inducing lymphocyte cell death(52) and for mediating the chemotactic effects of TSP1 on vascular smooth muscle cells.(53) However, a CD47-binding fragment from TSP1 does not inhibit EC proliferation.(50) Furthermore, in preliminary studies we have found that TSP2 is able to inhibit proliferation of CD47-null MSC as effectively as WT cells (K. D. Hankenson, M. Johanssen, E. Brown, and P. Bornstein, unpublished data, 2001).

Instead of binding to a cell surface receptor, TSP2 may function as an inhibitor of MSC growth by interfering with the binding and subsequent activation of a growth factor receptor by another ligand. Both TSP1(54) and TSP2 (K. D. Hankenson and P. Bornstein, unpublished results, 2001) bind to bFGF and could, in this manner, disrupt normal FGF signaling. In preliminary studies, TSP2-null MSCs show increased activation of the mitogen-activated protein kinases (MAPK) ERK1 and ERK2, a finding that suggests that growth factor signaling is enhanced in the absence of TSP2. Growth factors such as bFGF increase the level of G1 cyclins (D and E), which promote progression through the G1/S checkpoint of the cell cycle.(55)

Entrance into S phase is mediated also by the down-regulation of CDK inhibitors such as p21 and p27.(56) Levels of p27 increase with cell density or during mitogen deprivation and prevent the formation of active CDK complexes with cyclins D and E.(57) Preliminary results from our laboratory show that cells expressing TSP2 have increased levels of p27 (K. D. Hankenson and P. Bornstein, unpublished results, 2001). An increase in active MAPK leads to a loss of p27(58); thus, we hypothesize that TSP2 may up-regulate cyclin inhibitors by reducing the activation of MAPK. In this manner, TSP2 may play a role in regulating in vitro cell density. An initial study with TSP2-null skin fibroblasts indicated that they show increased cell density relative to WT cells.(32) As well, the level of TSP2 messenger RNA (mRNA) is decreased in response to elevated levels of c-myb, a proto-oncogene that is increased during density-independent cell growth.(59) Experiments with immortalized WT and TSP2-null cells further support a role for TSP2 in regulating cell density because TSP2-null cells are able to achieve a much higher cell number per plate after WT cells have experienced density-dependent quiescence (K. D. Hankenson and P. Bornstein, unpublished results, 2001).

Based on the evidence presented in this study, TSP2 appears to act as a novel extracellular regulator of MSC cell cycle progression. We hypothesize that TSP2 functions as an autocrine inhibitor of proliferation in the marrow milieu and maintains osteoprogenitor cells in a quiescent state both in vivo and in vitro. A determination of the expression levels of TSP2 in vivo during conditions that are known to affect marrow osteoblastogenesis negatively or positively, such as fracture healing, mechanical stress, or estrogen depletion, will be helpful in gaining a better understanding of TSP2 and MSC biology.


The authors thank Kevin Otipoby and Jessica Hamerman for providing advice and reagents for immunophenotyping. Jennifer Tullis, Qian Zhang, and Emily Stainbrook provided invaluable technical assistance. This work was supported by National Institutes of Health (NIH) grants HL18645 and AR45418 (to P.B.). K.D.H. was supported by NIH Training grant DE07063 and a special emphasis research career award from the National Center for Research Resources (RR0161).