Attenuation of WNT signaling by DKK-1 and -2 regulates BMP2-induced osteoblast differentiation and expression of OPG, RANKL and M-CSF

To evaluate the role of mouse Dkk proteins in cancer-associated focal bone lesions in mice, we generated four different sublines of mouse plasmacytoma MOPC315.4 that stably expressed Dkk1, -2, -3 or -4. Tumor cells were transfected with retrovirus that contained a bi-cistronic expression vector encoding one of the Dkk proteins in conjunction with the reporter, enhanced GFP. The reporter was used for enriching GFP-expressing cells by flow sorting, which increased, in the case of Dkk1, the fraction of GFP⁺ cells from 70% to 90%. Similar enrichments were achieved with Dkk2, -3 and -4. All four Dkk proteins were readily detected by Western blotting of lysates of GFP⁺ tumor cells using either Ab to the FLAG epitope (Fig. 1A top) or specific Ab to Dkk1, -2, -3 or -4 (Fig. 1A center). Western blotting with Ab to β-actin and GFP showed that parental MOPC315.4 cells or cells transfected with retrovirus expressing only GFP did not express any detectable Dkk protein (Fig. 1A bottom). Real-time PCR of Dkk1, -2, -3 and -4 mRNA levels confirmed the negative immunoblotting data presented in Figure 1A (not shown).

Western analysis of cell culture supernatants from the above-described MOPC315.4 transfectants demonstrated that all four Dkk proteins were secreted into the extracellular milieu (Fig. 1B). All four Dkk's were readily detected by antibody to the His₆ tag (Fig. 1B top), FLAG (results not shown) or specific epitopes on the Dkk1, -2, -3 or -4 proteins (Fig. 1B bottom). MOPC315.4 cells left untreated or transfected with retrovirus expressing only GFP did not secrete Dickkopf protein (Fig. 1B, lanes labeled Parental and Control, respectively). Just like human DKK secreted by transient 293T human embryonic kidney cell transfectants [33], the secreted mouse Dkk proteins appeared to undergo post-translational modification including proteolytic cleavage and glycosylation, which produced distinct, highly reproducible bands on the immunoblots (Fig. 1B and data not shown).

To investigate whether the secreted Dkk proteins functioned as inhibitors of Wnt/β-catenin signaling, we performed the β-catenin stabilization assay in L cells [34, 35]. Stimulation of L cells with Wnt3a-CM (obtained from Wnt3a-transfected L cells) resulted in a marked increase of β-catenin compared to L cells stimulated with L-CM (obtained from mock-transfected L cells), as shown in Figure 1C (compare lanes 3–4 with lanes 1–2). The Wnt3a-dependent accumulation of β-catenin was reduced in cells treated with Dkk1- or Dkk2-CM (obtained from transfected MOPC315.4 cells) in a dose-dependent manner (Fig. 1C and 1D). Dkk3-CM was ineffective, as expected, whereas Dkk4-CM, which was marginally effective in some experiments described below, was also ineffective in these experiments (Fig. 1D). These results clearly demonstrated that Dkk1 and -2 secreted by mouse plasmacytoma cells functioned as canonical Wnt/β-catenin inhibitor.

To investigate the role of Dkk proteins in Wnt-induced osteoblast differentiation in vitro, we took advantage of cell lines C3H10T1/2 and C2C12. The former is a mouse mesenchymal stem cell with differentiation potential along several lineages including osteoprogenitors. The latter is a primitive mouse myoblast that can undergo differentiation to preosteoblast-like cells. In both cell lines, osteoblastic differentiation can be monitored by expression of ALP and OCN, two representatives of early and late markers of osteoblast differentiation, respectively. Treatment with Wnt3a led to a 5-fold increase in ALP activity in C3H10T1/2 cells (Fig. 2A right, top bar), confirming a previous result under similar conditions [10]. The ALP increase in C2C12 cells was not as pronounced (2-fold) but clearly present (Fig. 2A left, top bar).

Treatment of both cell lines with Dkk1 and -2 resulted in a significant drop in Wnt3a-induced ALP (Fig. 2A). Unlike C2C12 cells, treatment of C3H10T1/2 cells with Dkk4 led to a small but highly significant decrease in ALP. This indicated that Dkk4 sometimes exhibited Wnt-inhibiting activity, as mentioned above. Dkk3 was ineffective in both cell lines, as expected (Fig. 2A). The ALP-inhibiting effect of Dkk1 and -2 was confirmed at the level of gene transcription using real-time RT-PCR (Fig. 2B). Just as in case of ALP protein, Dkk1 was a potent inhibitor in both cell lines, whereas Dkk2's inhibitory activity was more convincing in C2C12 than in C3H10T1/2 cells. In contrast to Dkk3 and -4 (results not shown), Dkk1 and -2 inhibited the Wnt3a-dependent induction of OCN mRNA (Fig. 2C), with Dkk2 being once again more potent in C2C12 than C3H10T1/2 cells. These quantitative differences aside, the results suggested that Dkk1 and -2 inhibit Wnt/β-catenin-dependent osteoblast differentiation.

The developmental cues emanating from BMP signaling must be integrated with inputs from Wnt and other signal transduction pathways to properly orchestrate the network of transcription factors that govern osteoblast differentiation. Based on this consideration, we assessed osteoblastogenesis under conditions of ongoing BMP or Wnt/BMP signaling further modulated by the BMP inhibitor, Noggin, and the Wnt inhibitor, Dickkopf. Similar to previous reports on BMP2-induced differentiation of preosteoblasts [36, 37], treatment of C2C12 cells with BMP2 led to a substantial induction of ALP protein (~150-fold, Fig. 3A left, top bar) and ALP mRNA (~6000-fold, Fig. 3A right, top bar). This response was abrogated by Noggin (2^nd bar from bottom), indicating that ALP induction was dependent on BMP signaling. Surprisingly, we also found that in BMP2-stimulated cells, addition of Wnt3a resulted in a significant down-regulation of ALP (Fig. 3A, 2^nd bar from top), even though Wnt3a weakly induced ALP in absence of BMP2 (Fig. 2A). Treatment of BMP2-stimulated C2C12 cells with Dkk1 led to a more than two-fold increase in ALP protein (Fig. 3A left, 3^rd bar from top), not a decrease as reported elsewhere [27]. The increase was blunted by Wnt3a (Fig. 3A left, 4th bar), indicating that the Dkk1 effect was indeed dependent upon Wnt signaling, not caused by some unrelated property of this inhibitor. Dkk2 shared the above-described activities of Dkk1, while Dkk3 and -4 were ineffective (data not shown). These results indicated that in BMP2-stimulated osteoprogenitors the Wnt3a-dependent inhibition of ALP is partially abrogated by Dkk1 and -2.

Because transcription factors Cbfa1, Osterix and ATF4 are instrumental for the cell fate decision that commits mesenchymal progenitors to undergo differentiation to osteoblasts [30, 31, 38], we sought to determine whether Wnt3a/BMP2 signaling influences the expression of these transcription factors in differentiating C2C12 cells. Using real-time RT-PCR, we found that BMP2 caused but a modest (2-fold) increase in Cbfa1 mRNA (Fig. 3B left, top bar), yet a substantial (~60-fold) elevation in Osterix mRNAs (Fig. 3B right, top bar). ATF4 message was unchanged (data not shown). The induction of the Cbfa1 and Osterix genes was inhibited by Noggin (2^nd bar from bottom), indicating that the transcriptional activation relied on BMP2 signaling. Similar to ALP, the expression of Osterix, but not Cbfa1, was down regulated by Wnt3a (Fig. 3B right, 2^nd bar from top). Treatment of BMP2- or BMP2/Wnt3a-stimulated cells with Dkk1 resulted in both cases in a marked elevation of Osterix message (3^rd bar and 4^th bar, respectively), further indicating that Wnt signaling is a suppressor of Osterix. Dkk2 shared the above-described activities of Dkk1, while Dkk3 and -4 were ineffective (data not shown). These findings suggested that Wnt/β-catenin inhibits BMP2-dependent induction of Osterix in osteoprogenitors.

Osterix expression in osteoblasts has been recently shown to be negatively regulated by p53 [32]. We therefore speculated that the BMP2- or Wnt3a-dependent up- or down-regulation of Osterix may be accompanied by corresponding changes in p53 levels: decrease of p53 in case of BMP2 treatment and increase of p53 in case of treatment with Wnt3a. To evaluate this, we determined p53 in C2C12 cells using Western blotting. Lysate from untreated cells contained levels of p53 that were markedly reduced upon treatment with BMP2 (Fig. 3C, lanes 1 and 2). Treatment of cells with Wnt3a, which left the p53 baseline essentially unaffected (lane 3), restored p53 to near baseline levels in the presence of BMP2 (lane 4). Addition of Dkk1 or -2 did not alter the levels of p53 in Wnt3a-stimulated cells (compare lanes 7 and 11 with lane 3), but slightly attenuated the Wnt3a-dependent restoration of p53 under conditions of ongoing BMP2 signaling (compare lanes 8 and 12 with lane 4). Since p53 is a positive regulator of p21 expression, we next examined whether the BMP2-dependent loss of p53 was associated with the reduction of p21. Figure 3C (center) shows that this was the case. In fact, the changes of p21 closely matched those of p53 regardless of the specific treatment condition. These findings suggested that Wnt/β-catenin signaling by itself is not important for the regulation of p53/p21 in preosteoblasts. However, in cells undergoing BMP2 signaling, the Wnt/β-catenin pathway functions as a strong antagonist of the BMP-mediated suppression of p53.

Wnt3a/β-catenin signaling has been reported to induce OPG transcription in C3H10T1/2 cells [13], but this has not yet been extended to C2C12 cells, and the influence of Dkk proteins and ongoing BMP2 signaling on OPG expression are also unknown in both cell types. To that end, we found that treatment of C2C12 cells with Wnt3a resulted in a 6-fold induction of OPG mRNA (Fig. 4A, 2^nd bar from top). This induction was abrogated by Dkk1 (3^rd bar) and strongly inhibited by Dkk2 (4^th bar). Dkk3 and -4 were ineffective (not shown). Similar findings were obtained with C3H10T1/2 cells (not shown). To further validate these results, we used ELISA to measure the amount of OPG protein secreted into the cell culture supernatant. There was a good match of OPG message and protein (Fig. 4B) for all four treatment conditions. The same result was obtained with C3H10T1/2 cells (not shown). We next measured OPG mRNA in BMP2-stimulated and BMP2/Wnt3a-costimulated C2C12 cells. BMP2 decreased OPG mRNA below the level seen in the control (Fig. 4C top, 2^nd bar). The decrease was abolished by Noggin (3^rd bar), but not by Dkk1 (4^th bar). Wnt3a partially restored the level of OPG mRNA held down by treatment of cells with BMP2 (Fig. 4C bottom, 2^nd bar). The restoration was abolished by Dkk1 (4^th bar), but not by Noggin (3^rd bar). These results demonstrated that OPG is regulated by both Wnt/β-catenin and BMP2 signaling. The Wnt pathway induces OPG, whereas the BMP2 pathway suppresses OPG. Interestingly, Dkk1 kept OPG low irrespective of the BMP2 signaling status (on or off). This pointed to a mechanism by which Dkk1 may promote osteolytic lesions in vivo: enhancement of RANKL-mediated osteoclastogenesis.

Wnt/β-catenin signaling has been demonstrated to down-regulate RANKL mRNA in MC3T3-E1 preosteoblast [14], but thus far this situation has not been studied in C2C12 or C3H10T1/2 cells. The influence of Dkk proteins on RANKL expression and the outcome of possible crosstalk of the Wnt, BMP2 and 1,25(OH)₂D₃ signaling pathways are also unknown. On this backdrop, we found that Wnt3a repressed RANKL mRNA in C2C12 cells (Fig. 5A) and C3H10T1/2 cells (results not shown). In both cell lines RANKL suppression was mitigated by Dkk1, whereas Dkk2, -3 and -4 and Noggin were ineffective (Fig. 5A and data not shown). Wnt3a also repressed 1,25(OH)₂D₃-induced RANKL mRNA expression in C2C12 cells (Fig. 5B), which was partially reversed by Dkk1 but not by Noggin (Fig. 5B), indicating involvement of the Wnt/β-catenin pathway. BMP2 increased RANKL mRNA in C2C12 cells (Fig. 5C), which was inhibited by Noggin, as expected. Importantly, because Wnt3a was unable to down regulate the BMP2-dependent induction of RANKL, the addition of Dkk1 to this system was inconsequential (Fig. 5C). This indicated that in contrast to 1,25(OH)₂D₃-stimulated and untreated C2C12 cells, cells undergoing BMP signaling become unresponsive to Wnt/β-catenin's inhibiting effects on RANKL expression.

Osteoblast-produced M-CSF, a positive regulator of osteoblast-dependent osteoclast development, might be influenced by Wnt or BMP2 signaling in mesenchymal stem cells undergoing osteoblastic differentiation. To investigate this, we used C2C12 as the model system and real-time PCR as the measurement tool. We found that Wnt3a repressed M-CSF mRNA compared to the control (Fig. 6A, 2^nd bar from top). Unlike Noggin, Dkk1 restored the M-CSF message, further indicating that M-CSF is down regulated by Wnt/β-catenin signaling in this cell type. ELISA of C2C12 cell culture supernatant confirmed these findings at the level of secreted, soluble M-CSF protein (Fig. 6B). We next evaluated M-CSF production in C2C12 cells undergoing BMP2 and Wnt3a/BMP2 signaling further modulated by Noggin and Dkk1. BMP2 enhanced M-CSF by a factor of two (Fig. 6C top). Noggin, but not Dkk1, blocked the response. Wnt3a repressed the BMP2-induced increase of M-CSF in a manner that was reversed by Dkk1 but not Noggin (Fig. 6C bottom), confirming that the Wnt/β-catenin pathway is a negative regulator of M-CSF. FACS analysis of the same cell preparation evaluated in panel C demonstrated a good match of the secreted and membrane-associated forms of M-CSF (Fig. 6D). These results demonstrated that M-CSF expression in osteoblast progenitors is up and down regulated by BMP and Wnt/β-catenin signaling, respectively. Dkk1 stabilized M-CSF expression levels in presence and absence of BMP2, suggesting yet another mechanism by which Dkk1 activates osteoclast development in vivo.

Mouse plasmacytoma MOPC315.4 [50], a generous gift from Dr. B. Bogen (University of Oslo, Oslo, Norway), was cultured in RPMI 1640 containing 10% fetal bovine serum (FBS), 2 mM glutamine, 50 μM 2-mercaptoethanol, and antibiotics (100 U/ml penicillin, 100 μg/ml streptomycin). Plat-E cells [51] were kindly provided by Dr. T. Kitamura (Institute of Medical Science, University of Tokyo, Tokyo, Japan) and maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% FBS, 2 mM glutamine, 10 μg/ml blasticidin, 1 μg/ml puromycin, and antibiotics. C2C12 and C3H10T1/2 are mesenchymal stem cells obtained from ATCC (Manassas, VA). Wnt3a-expressing L cells (L-Wnt3a) and the corresponding parental cell line, L, were also purchased from ATCC. Cells were maintained in α-minimal essential medium in the case of C3H10T1/2 cells or DMEM in the case of C2C12, L-Wnt3a and L cells. Both media were supplemented with 10% FBS, 2 mM glutamine, and antibiotics. All cell culture media and supplements were obtained from Invitrogen (Carlsbad, CA). All cells were cultured in a humidified atmosphere at 37°C and 5% CO₂.

The cDNAs encoding Dkk1 and -4 were cloned by PCR, using a mouse embryonic cDNA library (Open Biosystems, Huntsville, AL) as template. Dkk2 and -3 cDNAs were obtained by reverse transcriptase-PCR, using total RNA from mouse uterus and brain, respectively, as template. All four cDNAs were modified by adding C-terminal FLAG or His₆ epitope tags, sequenced, and subcloned into the retroviral expression vector pMY-IRES-EGFP [52] kindly provided by Dr. T. Kitamura (Institute of Medical Science, University of Tokyo). pMY-IRES-EGFP without Dkk insert was used as control. Retrovirus was produced by transient transfection of Plat-E cells. Briefly, 1.6 × 10⁶ cells were plated in DMEM in 60-mm cell culture dishes for 30 h prior to transfection with retrovirus. Using Lipofectamine 2000 (Invitrogen), cells were transfected with 8 μg of one of the four Dkk expression vectors or the "empty" GFP control vector. The viral supernatant was harvested 16 h later and filter-sterilized. For retroviral infection, 5 × 10⁵ MOPC315.4 cells were cultured for 16 h in a 6-well plate that contained equal volumes of cell culture medium and retroviral supernatant supplemented with 10 μg/ml polybrene (Sigma, St. Louis, MO). Cells were washed and cultured for 3 days, followed by isolation of virally infected, green fluorescent protein (GFP) expressing cells using a FACSCalibur and the CellQuest software (Becton Dickinson, Franklin Lakes, NJ).

Medium conditioned with Dkk1, -2, -3 or -4 (hereafter referred to Dkk-CM) was obtained from flow-sorted GFP⁺ plasmacytoma cells. Briefly, 7.5 × 10⁵ cells were seeded in 75 cm² flasks containing 25 ml RPMI 1640 supplemented with 10% FBS. After 3 days in culture, supernatant was harvested, filter-sterilized, and stored at 4°C. Medium conditioned with Wnt3a (Wnt3a-CM) was prepared from commercially available L-Wnt3a cells [34]; 1 × 10⁶ cells were seeded in 100-mm plates containing 10 ml DMEM supplemented with 10% FBS. After cell culture for 4 days, the supernatant was harvested, sterilized, and stored at 4°C. Medium conditioned with flow-sorted GFP⁺Dkk^- plasmacytoma cells (M-CM) and parental L cells (L-CM) were used as controls for experiments with Dkk1-4 and Wnt3a, respectively.

Cells were incubated on ice for 20 min in CelLytic-MT (Sigma) supplemented with a proteinase inhibitor cocktail (Roche, Indianapolis, IN) followed by centrifugation at 12,000 × g for 20 min to prepare whole cell lysate. Protein concentration was determined with the help of the Bio-Rad Protein assay using BSA as standard (Bio-Rad Laboratories, Hercules, CA). Protein (10 μg lysate or 10 μl CM) was incubated for 10 min at 70°C in sample buffer containing reducing reagents; fractionated by SDS-PAGE using 4–12% NuPAGE gels in MES buffer (Invitrogen); transferred to a polivinylidene fluoride membrane (Invitrogen); blocked with StartingBlock T20 buffer (Pierce, Rockford, IL); labeled with un-conjugated primary antibody followed by horseradish peroxidase-conjugated secondary antibody; and visualized with the enhanced chemiluminescence detection system (GE Health Sciences, Piscataway, NJ). The following antibodies/-sera were used: mouse monoclonal to β-actin or FLAG (Sigma), GFP (MBL, Woburn, MA), His₆ (Invitrogen), or p21 (BD PharMingen); rat monoclonal to Dkk1 or -3 and goat anti-Dkk2 or -4 (R&D Systems, Minneapolis, MN); rabbit anti-p53 (Vector Laboratories, Burlingame, CA).

Wnt3a-dependent stabilization of β-catenin was determined as described elsewhere [35]. L cells (4 × 10⁵) were seeded in a 6-well plate, incubated for 24 h, washed, and transferred to 1 ml Dkk-CM or M-CM (control). Following incubation for 3 h, Wnt3a-CM or L-CM was added for 3 h. Cells were lysed in 50 μl lysis buffer. β-catenin was determined by immunoblotting using 10 μg protein as sample and mouse monoclonal anti-β-catenin from BD transduction Laboratories as detection tool. β-catenin levels were quantified by densitometry using NIH Image software. β-catenin expression was normalized by comparison to the housekeeping protein, β-actin.

Cells were seeded at 2 × 10⁴ cm^-2 one day prior to differentiation induction. Cell culture medium was replaced with α-minimal essential medium supplemented with 50 μg/ml ascorbic acid and Dkk-CM, BMP inhibitor (100 ng/ml mouse Noggin [R&D Systems]) or no inhibitor (M-CM). Following pre-incubation for 3 h, cell differentiation was induced by adding Wnt3a-CM and/or L-CM that either contained 100 ng/ml recombinant human BMP2 (R&D Systems) or 10 nM 1,25-dihydroxyvitamine D₃ (1,25(OH)₂D₃, Sigma). Cells were cultured for 3 days (if longer, the differentiation-inducing medium was replaced after 3 days). In some experiments, differentiation occurred in presence of Dkk-CM and/or 100 ng/ml Noggin. Dkk-CM and Wnt3a-CM were added to a final concentration of 20% (v/v), which had been found in pilot studies to result in optimal assay performance in our hands (results not shown).

Total RNA was isolated using the RNeasy mini kit (Qiagen, Valencia, CA) supplemented with DNase. First-strand cDNA synthesis was performed using the Superscript III kit (Invitrogen) with 1 μg RNA as template. Real-time PCR was carried out with the ABI 7500 cycler (Applied Biosystems, Foster City, CA). Briefly, 12.5 μl SYBR green PCR master mix (Applied Biosystems) and 5 μl cDNA (five- to ten-fold dilutions of original sample) were added to 200 nM forward and reverse PCR primers in 96-well microplates. Following thorough heat denaturation for 10 min at 95°C, cDNA was amplified in 40 two-step cycles using the following temperature protocol: 95°C for 15 s and 60°C for 60 s. PCR primer sequences and length of amplicons are listed in Table 1. PCR primers were designed by Roche Applied Science as part of their Universal ProbeLibrary. All PCR reactions were performed in triplicates. Results were normalized using β-actin message as reference.

Cell culture supernatant was harvested 3 days after initiation of osteoblast differentiation as described above. OPG, RANKL and M-CSF were determined by sandwich ELISA, using the respective Duo-Sets from R&D Systems.

Cells (1 × 10⁶) were incubated with 0.25 μg biotinylated monoclonal rat antibody to mouse CSF-1 (BD PharMingen) for 1 h at 4°C, followed by incubation with phycoerythin-labeled avidin (eBioscience, San Diego, CA). After washing in phosphate-buffered saline, cells were analyzed on a FACSCalibur using the CellQuest software for data evaluation. Biotinylated rat IgG2a (eBioscience) was used as control for unspecific antibody binding.

Mean values and standard deviations (SD) were calculated and compared using ANOVA followed by Dunnett's multiple comparison t-test. p values less than 0.05 were considered significant.