Comparison of TGF-β/BMP Pathways Signaled by Demineralized Bone Powder and BMP-2 in Human Dermal Fibroblasts


  • Shuanhu Zhou,

    1. Department of Orthopedic Surgery, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts, USA
    Search for more papers by this author
  • Julie Glowacki PhD,

    Corresponding author
    1. Department of Orthopedic Surgery, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts, USA
    • Orthopedic Research, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115, USA
    Search for more papers by this author
  • Karen E Yates

    1. Department of Orthopedic Surgery, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts, USA
    Search for more papers by this author

  • The authors have no conflict of interest


Demineralized bone induces chondrogenic differentiation of human dermal fibroblasts in vitro. Analyses of signaling gene expression showed that DBP and BMP-2 regulate common and distinct pathways. Although BMP-2 was originally isolated as a putative active factor in DBP, rhBMP-2 and DBP do not affect all the same genes or in the same ways.

Introduction: Demineralized bone powder (DBP) induces chondrogenic differentiation of human dermal fibroblasts (hDFs) in 3D culture, but the initiating mechanisms have not been identified. We tested the hypotheses that DBP would affect expression of signaling genes and that DBP's effects would differ from the effects of bone morphogenetic proteins (BMPs).

Materials and Methods: A chondroinduction model was used in which hDFs were cultured with and without DBP in a porous collagen sponge. BMP-2 was delivered in a square of absorbable collagen felt inserted into a collagen sponge. Total RNA was isolated after 3 days of culture, a time that precedes expression of the chondrocyte phenotype. Gene expression was evaluated with two targeted macroarray screens. Effects of DBP and rhBMP-2 were compared by macroarray, RT-PCR, and Northern hybridization analysis of selected genes in the transforming growth factor (TGF)-β/BMP signaling pathways.

Results: By macroarray analysis of 16 signal transduction pathways, the following pathways were modulated in hDFs by DBP: TGF-β, insulin/LDL, hedgehog, PI3 kinase/AKT, NF-κB, androgen, retinoic acid, and NFAT. There was convergence and divergence in DBP and rhBMP-2 regulation of genes in the TGF-β/BMP signaling pathway. Smad target genes were the predominant group of DBP- or rhBMP-2-regulated genes. Several genes (IGF-BP3, ID2, and ID3) showed similar responses (increased expression) to DBP and rhBMP-2. In contrast, many of the genes that were greatly upregulated by DBP (TGFBIig-h3, Col3A1, TIMP1, p21/Waf1/Cip1) were barely affected by rhBMP-2.

Conclusion: These findings indicate that multiple signaling pathways are regulated in fibroblasts by DBP, that one of the major pathways involves Smad target genes, and that DBP and rhBMP-2 elicit different gene expression responses in hDFs. Although BMP-2 was originally isolated as a putative inductive factor in DBP, rhBMP-2 and DBP do not affect all the same genes or in the same ways.


Demineralized bone implants are being used successfully in many types of craniomaxillofacial, orthopedic, periodontal, and hand reconstruction procedures.(1–3) The sequential cellular changes in response to implants of demineralized bone materials include chemotaxis and attachment of progenitor cells to the matrix; proliferation and differentiation of progenitor cells into chondrocytes; sequential chondrogenesis, cartilage mineralization, vascularization, and resorption of the induced cartilage; and ultimately osteogenesis and marrow formation.(4) Identification of the genes regulated by demineralized bone powder (DBP) in postnatal target cells should provide insight into the molecular mechanism of osteo/chondrogenesis induced by DBP.

Urist et al. proposed that bioactive proteins in demineralized bone matrix were responsible for inducing cartilage and bone formation and named these putative activities bone morphogenetic proteins (BMPs).(5) Since that time, BMPs have been isolated from matrix or prepared by recombinant DNA technology. They are known to induce various mesenchymal cells lines to differentiate to osteoblast and chondrocyte lineage cells.(6), (7) BMP-2 is a member of the transforming growth factor (TGF)-β superfamily and regulates endochondral bone formation.(8–11) Recombinant human BMP-2 (rhBMP-2) is being used clinically for spine fusions.(12) Therapeutic rhBMP-2 at high concentration is delivered through an absorbable collagen felt carrier(12) that is necessary for optimal activity of BMP-2 in skeletal induction.(13)

In previous studies, we developed a DBP/collagen sponge system that optimized interactions between particles of demineralized bone and target cells in culture.(14), (15) In that system, DBP induces chondrogenesis in human dermal fibroblasts (hDFs), as shown by histochemical, biochemical, and molecular markers of cartilage.(14), (16–19) When DBP/collagen sponges are implanted subcutaneously in animals, chondroinduction occurs vicinal to the particles of DBP.(15)

The relationship between BMP-2 and skeletal induction by DBP is unknown. In this study, we tested the hypotheses that DBP would affect genes in TGF-β/BMP signaling pathways and that DBP's effects would be different from the effects of BMPs. It is possible to compare cellular events elicited by DBP or BMP-2 in the 3D system because the BMP-2 is marketed with a collagen felt that can be inserted into the porous collagen sponges previously used as a support for DBP. We used macroarray analyses to identify pathways that are modulated by DBP and to make comparisons with rhBMP-2. RT-PCR and Northern hybridization were used to substantiate the macroarray findings.


Collagen sponges

Porous 3D collagen sponges (8 mm diameter) were prepared from pepsin-digested bovine collagen.(14) In brief, 250 μl of 0.5% collagen solution (Cellagen PC-5; ICN Biomedicals, Costa Mesa, CA, USA) was neutralized with 1 M HEPES (pH 7.4) and 1 M NaHCO3, poured into a mold, frozen, lyophilized, and irradiated with UV light. Control sponges consisted of a single layer (1.5 mm) of collagen. DBP was prepared from rat long bones.(20) Bilaminate collagen sponges were prepared by placing a spacer of moistened paper between two layers (0.7 mm each) of collagen. After irradiation of the sponges, DBP (3 mg) or rhBMP-2 was inserted between the two layers. The amount of rhBMP-2 (4 μl of 1.5 mg/ml rhBMP-2 in 0.1% bovine serum albumin [BSA]/PBS solution, total dose is 6 μg/sponge) was recommended by the manufacturer (Wyeth, Cambridge, MA, USA) in proportion to clinical use. It was delivered in a 4 × 4 × 2-mm3 piece of absorbable collagen scaffold (ACS) felt supplied by the manufacturer (Wyeth), with 0.1% BSA delivered by ACS as a control for rhBMP-2. Sponges were transferred to seeding chambers.(14)

Cells and culture conditions

Skin explants were discarded material (neonatal foreskin) obtained under approved institutional protocols.(14) Primary human dermal fibroblast cultures were established by outgrowth from minced tissue (1-mm3 pieces) and were expanded in vitro with DMEM (Invitrogen, Carlsbad, CA, USA), containing 10% FBS (Invitrogen) and antibiotics, to passage 12 before use. Suspensions of hDFs were seeded onto DBP/collagen (n = 13) or rhBMP-2/ACS/collagen (n = 4) and respective control collagen sponges (n = 13) or BSA/ACS/collagen (n = 4) contained in seeding chambers. The cells (106 cells in 50 μl/sponge) were deposited directly on top of the sponge. The seeded sponges were placed into a humidified chamber at 37°C with 5% CO2 in air for 1 h. An additional 50 μl of culture medium was added to each seeded sponge, and the sponges were returned to the incubator for 2 h for hydration. Three milliliters of culture medium was added, and the seeding chambers were transferred to 12-well tissue culture plates and tilted to ensure that no air bubbles remained beneath the sponges. The sponges were removed from the seeding chambers after 3 days of culture.

Histological analysis

One sponge from each group was fixed in 2% paraformaldehyde, 0.1 M cacodylate buffer, pH 7.4, at 4°C for 24 h. After being rinsed in 0.1 M cacodylate buffer, they were embedded in glycolmethacrylate (JB-4; Polysciences, Warrington, PA, USA). Cross-sections (10 μm thick) were cut in the central region of the sponges and were stained with 0.5% toluidine blue-O (pH 4.0; Fisher Scientific, Pittsburgh, PA, USA).

RNA isolation

Total RNA was isolated from pooled sponges by homogenizing in Trizol reagent (Invitrogen) according to the manufacturer's instructions.(17) RNA quality was evaluated by absorbance readings at 260 and 280 nm and by ethidium bromide staining of RNA separated by electrophoresis on formaldehyde-agarose gels.

cDNA macroarray analysis

Array design

On the GEArray Q series Human Signal Transduction Pathway Finder nylon macroarrays (catalog HS-008; SuperArray, Bethesda, MD, USA), cDNA features of 300–600 bp in length are spotted in quadruplicate in a 1-mm-diameter “tetrad” and are arranged in 8 columns and 14 rows. There are 96 experimental features per array. The six control features consist of plasmid (pUC18), blank (no DNA), and four potential normalization features (GAPDH, peptidylprolyl isomerase A [PPIA], ribosomal protein L13a [RPL13A], and β-actin [ACTB]).

Hybridization procedures

Aliquots (3 μg) of each pooled RNA were used for cDNA synthesis. Each sample was incubated for 90 minutes at 42°C in 10-μl reactions containing 30 μCi α-[32P]-dCTP, 200 U Moloney murine leukemia virus (MMLV) reverse transcriptase (Promega, Madison, WI, USA), and primers and reaction buffers provided by the manufacturer. Two microliters of Stop Solution and 2 μl of 10× denaturing solution (SuperArray) were added, and the labeled probes were denatured by incubation at 68°C for 20 minutes. Hybridization and washing steps were carried out in the plastic vials provided by the manufacturer (SuperArray), which were placed inside glass hybridization tubes. The nylon macroarrays were prehybridized in 2 ml of prewarmed GEAhyb Hybridization solution (SuperArray) containing 100 ng/ml salmon sperm DNA (Roche Molecular Biochemicals, Indianapolis, IN, USA) for 1 h at 60°C in a rotisserie-style hybridization oven. An aliquot of 0.75 ml prewarmed GEAhyb Hybridization buffer was added to each denatured probe and mixed by pipetting. The diluted probes were added to each vial, and hybridization was performed at 60°C. On the following day, two low-stringency (5 ml of 2× SSC, 0.5% SDS) and two high-stringency washes (5 ml of 0.1× SSC, 0.5% SDS) were performed for 5 minutes, each at 60°C, according to the manufacturer's instructions. An additional, high-stringency wash (10 minutes at 60°C) was performed after an initial autoradiograph showed high background signals.

Measurement data and specifications

Digital images were acquired from autoradiographs with an Epson 1200s Scanner with transparency adapter (EPSON TWAIN Presto! PageManager for EPSON version 4.20.03; EPSON, Long Beach, CA, USA). Images were acquired at 200 pixels/in and were formatted (squared, cropped, inverted, and saved in TIF format) with Paint Shop Pro version 5.01 (Jasc Software, Minneapolis, MN, USA). ScanAlyze 2.44 software (Stanford University, Palo Alto, CA, USA) was used to extract data. For each macroarray, the formatted image was loaded into channel 1 and channel 2 (gain = 1.0). A grid of spots (10 pixel diameter) was overlaid and centered over each feature tetrad. The average pixel intensity (API) within the spot was measured.

Evaluation of gene expression levels

GEArrayAnalyzer version 1.2 software was used to calculate gene expression levels for each feature. The usable range of API values with this instrumentation was 9000–55,000. This range was established from determining that the minimum (background) API value detected on the autoradiographs was ∼8000 and the maximum or plateau API was ∼65,000. API values that fell within those limits were accepted for analysis. API values that exceeded those limits were re-evaluated on another autoradiograph with a different duration of exposure. If re-evaluation was not possible because of background, gene expression levels were expressed as less than or greater than the estimated value. API values that were less than or equal to the minimum background value on an 18-h film exposure were classified as “absent calls.” API values for features that were located adjacent to highly expressed genes were excluded from analysis because of “bleeding” of the strong signals.

The data were transformed by subtraction of background (API of the pUC18 feature) and normalization for each exposure. Autoradiographs showed a wide range of expression levels that prevented accurate assessment of all features from a single exposure time. Therefore, multiple exposures were performed, and expression levels were calculated for each feature with the most appropriate of the four normalization controls (e.g., similar signal intensity) provided on the macroarray. Autoradiographs were inspected visually to confirm the most appropriate normalization control. Another requirement for selection of a normalization control was that the API value be within the established usable range.

Genes detected on the arrays were grouped according to expression level in control hDFs. Genes whose expression levels were measured on short (1 or 2 h) film exposures were classified as high expression (+++); 5-h film exposures were classified as low expression (++); and 18-h film exposures were classified as very low expression (+).

Quantitative data for gene expression levels in DBP/collagen and rhBMP-2/ACS/collagen sponges were expressed as the fold difference or percent difference relative to respective control (collagen or BSA/ACS/collagen). Numeric values for gene expression levels were calculated when signals on both the experimental and control arrays had an API within the user-defined range. If the API of a feature was below the lower limit of the usable range, changes in gene expression levels were presented qualitatively (up, down, or no change).

In the DBP experiment, 28% (27/96) of the genes on the Signal Transduction array were classified as present calls in control hDFs, and 25% (24/96) were classified as present calls in hDFs cultured with DBP. Three genes that were present at very low levels in control hDFs were below the limit of detection in hDFs cultured with DBP. Many genes were below the limit of detection in both control and hDFs cultured with DBP. Eighteen percent (17/96) of the genes on the Signal Transduction array were excluded from analysis (both control and DBP) because their features were adjacent to highly expressed genes.

In the TGF-β/BMP arrays, 33% (32/96) of the genes were classified as present calls in control hDFs, and 40% (38/96) were classified as present calls in hDFs cultured with DBP. Nine genes were excluded because their cDNAs were adjacent to highly expressed genes on the array. In the BMP-2 experiment, 35% (34/96) of genes were classified as present cells on the TGF-β/BMP array. Nine genes were excluded because their cDNAs were adjacent to highly expressed genes on the array, and 53 were classified as absent calls.


Two micrograms of total RNA was reverse-transcribed into cDNA with SuperScript II in 20 μl reaction volume (Invitrogen Life Technologies, Carlsbad, CA, USA) according to the manufacturer's instructions. One microliter of the cDNA was used in each PCR (21–25 cycles at 94°C for 1 minute, 60°C for 1 minute, and 72°C for 2 minutes) with human gene-specific primers. Primer sequences were as follows: TGFBI forward 5′-CATCCCAGACTCAGCCAAGA-3′, reverse 5′-GAGTTTCCAGGGTCTGTCCA-3′ (GenBank accession no. NM_000358, bases 1160–1419); IGFBP3 forward 5′-ATTGCGCTAAGGAGGACAAA-3′, reverse 5′-TGGGAACCTATAAAGGCAGG-3′ (HUMIBP3, bases 9926–10,226); Col3A1 forward 5′-GTTTTGCCCCGTATTATGGA-3′, reverse 5′-GGAAGTTCAGGATTGCCGTA-3′ (NM_000090, bases 157–550); and ID2 forward 5′-CGGTCTCGCCTTCCTCGCGGTC-3′, reverse 5′-CTTATTCAGCCACACAGTGCT-3′ (HUMID2X, bases 84–518). Primers for the housekeeping gene, GAPDH, were purchased from Clontech (Palo Alto, CA, USA). Published primer sets were used to amplify human p21/Waf1(21) or PAI-1.(22) Gene expression levels were measured by semiquantitative PCR. PCR products were quantitated by densitometry of scanned gels (measured by National Institutes of Health Image J Software) and expressed as the percent increase for DBP- or rhBMP-2-treated versus respective control after normalization to GAPDH.

Northern hybridization

Northern hybridizations were performed with modification as described previously.(17) In brief, 10 μg total RNA was electrophoretically resolved in a 1% agarose/formaldehyde gel and blotted onto a positively charged nylon membrane (Boehringer Mannheim) by downward capillary transfer in 20× SSC buffer. Probes for human p21, Col3A1, TGFBI, ID2, PAI-1, and IGFBP3 sequences were generated by RT-PCR and were labeled with [32P]-ATP (random primed DNA labeling kit; Roche Molecular Biochemicals). Control 18S ribosomal RNA oligo probe (Ambion, Austin, TX, USA) was labeled with the DNA 5′-End Labeling Kit (Roche Diagnostics). The nylon membrane was prehybridized in formamide prehybridization/hybridization buffer (5× SSC, 5× Denhardt's, 1% SDS, 50% formamide, 0.1 mg/ml denatured salmon sperm DNA) for 3 h at 42°C. The hybridization was performed in that buffer overnight at 42°C. The membrane was washed twice by 2× SSC/0.1% SDS (5 minutes each), twice by 0.2× SSC/0.1% SDS (5 minutes each) at room temperature, and twice by 0.2× SSC/0.1% SDS (15 minutes each) at 42°C. Hybridization was visualized by autoradiography at −80°C with Kodak X-Omat Blue XB-1 film (Kodak, Rochester, NY, USA). Bound probes were removed by incubation at 65° in stripping solution that contained 50% formamide. The sequence for probe hybridization and removal was as follows: p21/Waf1/CIP1, Col3A1, probe removal; TGFBI; probe removal; ID2, probe removal, PAI-1, probe removal, IGFBP3, probe removal, 18S RNA. Autoradiographs were scanned with an Epson 1200s Scanner and transparency adapter, and the images were analyzed with Scion Image software. Pixel density was used to measure gene expression levels and was normalized to 18S rRNA signal.


Multiple signaling pathways modulated by DBP in hDFs

To assess which signaling pathways are affected during early gene regulation by DBP in hDFs, we surveyed changes in gene expression with a nylon cDNA macroarray. Control hDFs cultured in collagen sponges without DBP expressed marker genes in several signaling pathways (Table 1), of which fibonectin, IGFBP3, and fatty acid synthase were among the most highly expressed. As expected, DBP affected expression of marker genes in the TGF-β superfamily signaling pathway (upregulation of IGFBP3 and p21/Waf1/Cip1 and downregulation of p16ink4; Table 1). Expression of marker genes in several other signaling pathways was also affected by DBP, including insulin/LDL (downregulation of fatty acid synthase, ELAM1/E-selectin, and leptin), Hedgehog (downregulation of Wnt1, Forkhead box A2, and Patched 1), PI3 kinase/AKT (upregulation of Fibronectin), NFκB (downregulation of PECAM-1), androgen (downregulation of Prostate kallikrein 2), retinoic acid (downregulation of Engrailed homolog 1), and NFAT (downregulation of CD5). An early response gene, Egr-1, which was included on the array as a marker gene in the Creb, phospholipase C, and mitogenic pathways, was expressed at very low levels in control hDFs and was modestly increased by DBP. Marker genes for several signaling pathways on the array (Wnt, Stress, Jak-Stat, estrogen, calcium/protein kinase C) were not affected by DBP. One gene that was included on the array as a housekeeping control, β-actin, was found to be upregulated 1.9-fold by DBP.

Table Table 1.. Signaling Pathway Profile of HDFs in Collagen Sponges Without and With DBP (3 Days)8
original image

TGF-β/BMP signaling and target genes modulated by DBP in hDFs

The changes in TGF-β pathway genes that were identified in the signaling survey macroarray supported the hypothesis that TGF-β/BMP superfamily proteins within DBP may be involved in chondroinduction. To obtain a more detailed view of TGF-β superfamily signaling in hDFs exposed to DBP, we used a macroarray specific for these pathways. The most highly expressed TGF-β/BMP signaling and target genes in control hDFs included some Smad target genes (TGFBIig-h3, Col1A2, Col3A1, IGFBP3, TIMP1), as well as signaling proteins and transcription factors (v-Jun, RUNX2/cbfa1) (Table 2). Control hDFs also expressed some of the TGF-β superfamily receptors and Smad intracellular signaling proteins, albeit at relatively low levels (Table 2).

Table Table 2.. TGF-β/BMP Signaling and Target Gene Expression Profile of HDFs Treated With DBP or rhBMP-2 for 3 Days8
original image

The DBP-regulated TGF-β/BMP signaling pathway genes belonged to one main group, the Smad target genes (TGFBIig-h3, Col1A2, Col3A1, IGF-BP2, TIMP1, PAI-1, p21Waf1, ID2, ID3, ID4; Table 2). Some of the genes in other TGF-β signaling pathway groups (receptors, cytokines, and signaling molecules) were altered by DBP, but their expression levels in hDFs were low. Functional classification of those target genes revealed that DBP increased expression of several peptide differentiation factors (TGFBIig-h3, IGF-BP3, inhibin α), receptors (endoglin), and matrix proteins (COL1A2, COL3A1, fibronectin). Increases in other factors (BMP1, BMP antagonist-1, anti-mullerian hormone) and receptors (activin receptors I, IB, II, and II-like; BMP receptor IA) were detected but could not be quantified because of low levels in control cells. Decreases in some differentiation factor receptors (TGFR II and III, patched 1) were detected. It is notable that no changes in expression were found for Nodal (a differentiation factor) or BMP receptors IB or II. Several TGF-β/BMP signaling genes (RUNX1/AML1, RUNX2/cbfa1; Smad 2,3,4,5, and 9) were also not changed by DBP.

Because several genes were present on both the signal transduction and TGF-β/BMP arrays, they served as internal experimental controls (IGFBP3, p21/Waf1/Cip1, c-fos, Jun-B, c-myc, BMP-2, and BMP-4). Results for those genes were consistent in all experiments. There were no cases in which a gene identified as upregulated on one array was downregulated on the other or vice versa.

Histological analysis

Histological evaluation of cultured sponges 3 days after seeding with hDFs showed that cells were evenly distributed throughout the lattice of the porous collagen (Fig. 1). In the control collagen sponge, cells had a uniform spindle shape throughout the sponge (Fig. 1A). Sponges that contained particles of demineralized bone showed many hDFs attached to the particles (Fig. 1B). As expected, there was no evidence of metachromatic extracellular matrix at this early time. There was little evidence of migration of hDFs into the insert of ACS felt (Fig. 1C). In contrast, some hDFs were seen within the ACS felt insert that contained rhBMP-2 (Fig. 1D), but at a sparser density than in the collagen lattice of the sponge.

Figure FIG. 1..

Photomicrographs (toluidine blue-O stain; magnification, ×60) of cultured sponges 3 days after seeding with human dermal fibroblasts. (A) Plain porous collagen sponge has cells evenly distributed from surface to central region. (B) Porous collagen sponge with central packet of demineralized bone particles (D) has even distribution of cells on and between the particles. (C) Porous collagen sponge with central insert of control ACS felt has little cellular penetration of the felt. (D) Porous collagen sponge with central insert of ACS containing rhBMP-2 has some cellularity in the felt.20

TGF-β/BMP signaling and target genes modulated by rhBMP-2 in hDFs

After 3-day treatment of hDFs, rhBMP-2 upregulated TGFBIig-h3, COL1A2, IGFBP3, ID2, ID3, TGF1, endoglin, gremlin, ALK-2 and 3, inhibin βA, AMH, Smad 6 and 7 gene expression and downregulated PAI-1, Stat-1, p21/Waf1/Cip1, ID4, inhibin α, TGFBRII, and RUNX2/cbfa1 gene expression. The genes most affected by rhBMP-2 were ID3, IGFBP3, ID2, endoglin, and inhibin α (Table 2). BMP-2 moderately altered Smad target genes (TGFBIig-h3, PAI-1, Stat1, P21), receptors (TGFBRII, ALK-2, ALK-3), cytokines (inhibin βA, TGFβ1, AMH), and signaling molecules (gremlin, Smad 6 and 7).

Changes in TGFβ/BMP signaling and target genes induced by BMP-2 were compared with DBP-induced changes (Table 2, Fig. 2). First, there were very similar effects on some genes, such as increased expression of IGFBP3, ID2, and ID3. In contrast, some of the genes that were most dramatically increased by DBP, such as TGF-β-induced protein (1160%) and Col3A1 (1300%), were barely affected by rhBMP-2. Although there was concordance in stimulation of ID2 and ID3, ID4 was decreased 60% by DBP but only 10% by rhBMP-2. Both PAI-1 and TIMP1 were greatly increased by DBP, but PAI-1 was decreased by rhBMP-2 (30%), and TIMP was not at all changed. DBP increased cyclin-dependent kinase inhibitor p21/Waf1/Cip1 by 160%, whereas it was decreased (30%) by rhBMP-2. Cbfa1 was highly expressed in target hDFs but was moderately decreased by both DBP (10%) and rhBMP-2 (20%). Smad 6 and 7 were increased by only rhBMP-2. Smad 2–5 and 9 were very low-to-low in hDFs and were not modulated by either BMP-2 or DBP (Table 2).

Figure FIG. 2..

GComparison of DBP and rhBMP-2 effects on gene expression profiles of TGF-β/BMP signal transduction pathway in hDFs. Multiple exposures (1, 2, and 5 h and overnight) were performed and analysis was carried out with the most appropriate of the four normalization controls for each gene. The quantitative data of selected genes from macroarray are expressed as the percent increase or decrease for DBP/collagen relative to control collagen sponges or for rhBMP-2/ACS/collagen relative to control BSA/ACS/collagen sponges.20

RT-PCR analysis of selected genes regulated by DBP or rhBMP-2

Expression levels of selected genes that were shown to be altered by macroarray were evaluated by semiquantitative RT-PCR (Fig. 3A). These analyses showed similar changes as in the arrays but with different magnitudes (Fig. 3B). DBP upregulated Col3A1 (by 90%), TGFBIig-h3 (by 130% more than control), ID2 (by 100%), and IGFBP3 (by 130%). In contrast, rhBMP-2 upregulated TGFBIig-h3 (by 40%), ID2 (by 280%), and IGFBP3 (by 250%). Col3A1 was not significantly modulated by rhBMP-2.

Figure FIG. 3..

RT-PCR analysis of selected genes regulated by DBP or BMP-2. (A) Expression levels of selected genes that were shown to be altered by macroarray were evaluated by semiquantitative RT-PCR. Bands represent PCR products for samples treated with DBP or collagen control or with BMP-2 or BSA control. (B) Expression levels are presented as the percent difference for DBP- or rhBMP-2-treated vs. respective control after normalization to the housekeeping gene GAPDH.20

Northern hybridization analysis of selected genes regulated by DBP or rhBMP-2

Effects of DBP and rhBMP-2 on some of the target genes identified by macroarray were evaluated by Northern hybridization analysis (Fig. 4A). Representative genes that showed very low (p21/Waf1), low (ID2, PAI-1), or high levels of expression (Col3A1, TGFBIig-h3, IGFBP3) by macroarray assays were evaluated. Similar changes were found as in the arrays but with different magnitudes. DBP upregulated p21 by 97% more than control, Col3A1 by 178%, TGFBIig-h3 by 602%, PAI-1 by 470%, IGFBP3 by 350%, and no significant effects on ID2. RhBMP-2 increased TGFBIig-h3 by 173% more than control, ID2 by 234%, IGFBP3 by 120%, and not significantly modulated on p21, Col3A1, and PAI-1 (Fig. 4B).

Figure FIG. 4..

Northern hybridization of selected genes regulated by DBP or BMP-2. (A) Northern blot analysis for selected genes that were shown to be altered by macroarray. Northern blot was done as described in the Materials and Methods section. (B) Expression levels are presented as the percent difference of band density for DBP- or rhBMP-2-treated vs. respective control after normalization to 18S ribosomal RNA.20


Autograft, allograft, and synthetic bone graft substitute materials are important treatment options in reconstructive surgery, and understanding the biological effects of these materials is necessary for optimum implant design and use.(1), (2) Because many bioactive factors are present in bone matrix,(23) it is reasonable to expect that multiple signaling responses would be elicited by DBP. Our previous work showed that DBP induces expression of chondrocyte-specific genes for aggrecan and collagen type II in hDFs after 7 days of exposure.(16) Before full induction of the chondrocyte phenotype and chondrogenesis, 3 days of exposure to DBP alters expression of genes in various functional families.(17–19) A sampling of DBP-altered gene expression revealed that several signal transduction genes were elevated on day 3.(17), (19) One goal of this study was to further define the signal transduction pathways that may contribute to the initiating effects of DBP on human dermal fibroblasts (hDFs).

To identify DBP-induced changes in signal transduction genes, we first used a screening macroarray designed to monitor the activation of various signal transduction pathways. Several pathways were modified by DBP, including TGF-β, insulin/LDL, and PI3 kinase. The discovery of DBP effects on signaling genes in these pathways provides new information on the differentiation process. For example, DBP downregulated the insulin-pathway marker gene fatty acid synthase and the LDL-pathway marker gene ELAM1/E-selectin. Those findings imply that DBP may decrease fibroblasts' potential for adipogenesis, as they become committed to chondroinduction.

The TGF-β superfamily of cytokines influences a diverse range of normal cellular processes, such as cell adhesion, cell proliferation, differentiation, apoptosis, and secretion of extracellular matrix molecules.(24–29) TGF-β/BMPs are key autocrine and paracrine regulators of osteo/chondrogenesis.(7), (30–36) We found DBP effects on a few TGF-β/BMP superfamily signaling and target genes (COL3, ID2, IGF-BP3, TGFBIig-h3) by other methods.(17), (19) In this study, targeted analysis of DBP's effects on TGF-β/BMP superfamily signaling and target genes showed that hDFs expressed some receptors (TGFBRII, TGFBRIII, and AMHR2) and many intracellular signaling molecules (Smad 2, 3, 4, 5, 6, 7, 9) related to those agents. Smad-target genes were the ones most affected by DBP. In this 3D collagen sponge system, DBP dramatically upregulated TGFBIig-h3), Col1A2, Col3A1, IGFBP3, TIMP1, PAI-1, p21/Waf1/Cip1, ID2, and ID3, and downregulated ID4. Several of those genes (Col1A2, Col3A1, and TIMP1) are also known to be TGF-β target genes in hDFs cultured in monolayers.(37)

Several of the BMPs have been used in different carriers to promote bone formation in animal models.(13) Recently, a formulation of rhBMP-2 in an absorbable collagen felt has become available for use with tapered threaded fusion cages for interbody fusion.(12) To test whether DBP and BMP-2 affect similar signaling genes in the 3D in vitro model of induction, we compared the effects of DBP and rhBMP-2 on genes in the TGFβ/BMP signal pathway. Both DBP and rhBMP-2 upregulated ID2 and ID3, and downregulated ID4 gene expression in hDFs in 3D collagen sponges. The ID genes encode dominant negative helix-loop-helix (dnHLH) proteins that have been implicated in blocking both myogenesis and adipogenesis. They had been shown to be regulated by BMP-2 in embryonic stem cells,(38) in an immortalized human marrow stromal cell line,(39) and in C2C12 premyoblasts.(7) Concordance in regulation of those genes as well as Stat1 by DBP and rhBMP-2 suggests that the concentration of rhBMP-2 used (as recommended by the manufacturer) was apt.

Other Smad target genes that were significantly upregulated by DBP were not affected, minimally increased, or decreased by rhBMP-2. Although many of the receptors were stimulated by DBP, only ALK-2 and 3 were stimulated by rhBMP-2. Both agents downregulated TGFBRIII. Differences in receptor induction are likely to contribute to differences in effects. In these cells, rhBMP-2 was a stronger stimulus than DBP for a number of genes: TGFβ1, endoglin, gremlin, Smad 6, and Smad 7. It had been reported previously that rhBMP-2 enhances gene expression of TGF1 in human osteoblast-like cells,(40)gremlin in osteoblast-enriched cells from fetal rat calvariae,(41) and Smad 6 in various cells.(42)Smad 7 was previously shown to be rapidly induced by BMP, as well as by TGFβ, in C2C12 premyoblasts.(43) DBP increased cyclin-dependent kinase inhibitor p21/Waf1/Cip1, whereas it was not stimulated by rhBMP-2. Assessment of p21 and its regulation(44) frequently indicates status of normal cells in the cell cycle and has been correlated with differentiation. In fetal rat calvarial cells, p21 was increased on osteoblast differentiation.(45) The antiproliferative effects of p21 may play a role in osteoblast differentiation in other systems by other agents, such as pRb,(46) and fibroblast growth factor (FGF).(47) In another study of chondrogenesis by BMP-2 transduction, p21 was involved in the differentiation process.(48)

The different effects of DBP and BMP provide information about the mechanisms by which DBP acts on target cells. Scrutiny of the pattern of downstream targets indicates that DBP signals two paths of Smad activation, the one responsive to BMP as well as the one responsive to activin/TGFβ (Fig. 5), as recently classified by two reviews.(24), (49) That classification helped to explain how the many members in the TGF-β/BMP superfamily have such a broad array of biological activities with only a finite number of signaling receptors. In this study, DBP regulated many of the Smad target genes that are known to respond specifically to BMP through Smads 1, 5, and 8: namely, ID 2–4 and STAT1. Indeed, those genes were regulated by rhBMP-2 in this study. In addition, DBP regulated many of the Smad target genes that are known to respond specifically to activin/TGF-β through Smads 2 and 3: TGFBIig-h3, Col1A2, Col3A1, IGFBP3, TIMP1, PAI-1, p21/Waf1/Cip1, JUN, and TGF1. Thus, with respect to these signaling molecules, DBP affects genes that are targets of both subgroups of regulatory Smads. Because many bioactive factors besides BMP-2 are present in bone matrix, it is reasonable to expect that multiple responses would be elicited by DBP. In addition, control hDFs showed undetectable or very low expression of most TGF-β superfamily cytokines, such as BMPs, growth and differentiation factors (GDFs), and TGF-βs. Of those, only BMP-1, anti-Mullerian hormone, and inhibin α were upregulated by DBP. Combined, these results suggest that Smad target genes were modulated by various constituents in DBP.

Figure FIG. 5..

A model of the effects of different families of ligands on Smads and their target genes. Activin/TGF-β members bind to receptors RA that activate AR Smads, whereas the BMP/GDF factors bind to receptors RB that activate BR Smads.(24), (43) On the basis of Smad target gene changes, we conclude that DBP acts through both AR and BR Smads.20

Other recent data have given insights into the transcriptional networks controlled by BMP or TGF-β. In the pluripotent mesenchymal precursor cell line, C2C12, Runx2 expression was increased 2–6 h after exposure to either BMP-2 or TGF-β; after 3 days, however, only the BMP-2 treated cells expressed osteoblastic marker genes.(50) BMP-2-induced osteogenic differentiation is also mediated in part by Dlx-5, whose expression was suppressed by TGF-β.(51)

The differences between DBP and rhBMP-2 effects may, in part, shed light on the weak effects of the latter on normal human target cells. In the first of a series of comparative studies, it was shown that 18 of 19 samples of human bone marrow stromal cells did not increase alkaline phosphatase expression in response to rhBMP-2, in marked contrast to rat or mouse stromal cells.(52) In a subsequent study, it was reported that, although alkaline phosphatase was not induced by BMP-2 alone, other markers of osteoblast differentiation were induced, e.g., bone sialoprotein, osteopontin, BMP-2, and noggin.(53) There was a suggestion of an age effect because, when pretreated with dexamethasone, samples from the younger subjects showed stimulation of alkaline phosphatase by BMP-2 and those from older subjects did not.(54) It is also notable that milligram quantities of rhBMP-2 are required for clinical use, not in proportion to its low content in clinically effective demineralized bone products.(12)

In conclusion, this study indicates that multiple signaling pathways are involved in DBP's effects on hDFs. One of major pathways involves Smad target genes. Although BMP was originally isolated as the active factor in DBP, rhBMP-2 alone and DBP do not have the same effects on genes. The divergent and convergent actions of DBP and BMP-2 show that multiple members of the TGF-β superfamily, besides BMP-2, and other constituent factors may be involved in skeletal induction of hDFs by DBP.


We greatly appreciate help from Dr S Mizuno and KD Johnson in aspects of these experiments. Wyeth made a gift of the rhBMP-2 and ACS felt. This study was supported in part by National Institutes of Health Grant AR044873 to JG.