A Crosstalk Between Myeloma Cells and Marrow Stromal Cells Stimulates Production of DKK1 and Interleukin-6: A Potential Role in the Development of Lytic Bone Disease and Tumor Progression in Multiple Myeloma
William G. Gunn,
Center for Gene Therapy, Tulane University Health Sciences Center, New Orleans, Louisiana, USA
Multiple myeloma (MM) is a malignancy of antibody-secreting plasma cells. B-cell plasmacytomas stimulate bone resorption and angiogenesis, resulting in osteolytic lesions in the skeleton which persist upon successful treatment of the malignancy with chemotherapy. We found that an interaction between MM cells and mesenchymal stem cells (MSCs) from bone marrow stroma results in the formation and persistence of osteolytic bone lesions. It is known that MM cells activate osteoclast activity and secrete high levels of the Wnt inhibitor, Dickkopf-1, which prevents MSCs from differentiating into osteoblasts. We show that the Wnt signaling activator 6-bromoindirubin-3′-monoxime (BIO) releases MSCs from the osteoinhibitory effects of Dickkopf-1, whereas LiCl treatment does not. Additionally, we show that the >5-kDa fraction of MSC-conditioned medium promotes the proliferation of Dickkopf-1-secreting MM cells and that an interleukin-6 (IL-6)-neutralizing antibody blocks this effect. IL-6 expression levels were higher in undifferentiated MSCs than in MSCs treated with osteogenic medium, remained high in the presence of Dkk1, and were reduced by BIO treatment. Therefore, BIO treatment reduces the MSC-stimulated proliferation of MM cells and may enable MSCs to repair existing osteolytic lesions.
Tian et al.  recently demonstrated that Dkk1, a Wnt antagonist, is produced at high levels by myeloma cells from patients with osteolytic lesions and suggested that Dkk1 may inhibit differentiation of bone marrow stromal cells into osteoblasts. Earlier, Gregory et al.  had found that the human homolog of Dkk1 drives entry of mesenchymal stem cells (MSCs) into the cell cycle. Therefore, the results raised the possibility that Dkk1 may be involved in osteolytic lesion formation in multiple myeloma (MM), a debilitating and incurable malignancy of antibody-secreting plasma cells [3, 4]. Osteolytic lesions in MM are the major source of morbidity and occur where B-cell plasmacytomas stimulate bone resorption and angiogenesis . MSCs differentiate into osteoblasts at the lesion sites and attempt to repair the lesions by restoring the balance between bone production and resorption. However, adequate bone repair does not occur in myeloma patients , probably because of inhibition of repair by Dkk1. Previous work has shown that interactions between MSCs and MM cells support the proliferation of myeloma [3, 6–13]. We found that an alteration of the bone marrow microenvironment by MM cells explains the formation of osteolytic bone lesions. The MM cells secrete Dkk1, which prevents the MSCs from differentiating into osteoblasts, and the undifferentiated MSCs produce interleukin-6 (IL-6), which stimulates the proliferation of Dkk1-secreting MM cells. Finally, we show that use of a Dkk1-blocking agent, such as 6-bromoindirubin-3′-monoxime (BIO), can release MSCs from the MM proliferation-promoting cycle induced by Dkk1, which may also enable MSCs to repair existing osteolytic lesions.
Materials and Methods
For the preparation and culture of MSCs, we obtained bone marrow aspirates from healthy donors under a protocol approved by the institutional review board. The Tulane Center for Preparation and Distribution of Adult Stem Cells (New Orleans) supplied frozen vials. To recover MSCs, we added thawed vials containing 1 × 106 cells to 25 ml of complete culture medium (CCM) comprised of α-minimum essential medium (Invitrogen, Carlsbad, CA, http://www.invitrogen.com), 2 mM l-glutamine (Invitrogen), and 20% fetal bovine serum (FBS; Atlanta Biologicals, Norcross, GA, http://www.atlantabio.com) per 145-cm2 plate (Corning Life Sciences, Pittsburgh, http://www.corning.com/lifesciences). Our cell culture core facility selected batches of FBS for rapid growth of cells. For some experiments, we added 100 units/ml penicillin G and 100 μg/ml streptomycin to the CCM. After 1 day, we aspirated the medium, removing nonadherent and nonviable cells, then added fresh medium. Thereafter, we replaced the medium every 2 days. Four to 5 days after initial plating, we detached the cells by adding 0.25% (wt/vol) trypsin and 1 mM EDTA (Invitrogen) for 5 minutes at 37°C. We pelleted MSCs by centrifugation at 453g and then re-plated them at the indicated density. We expanded ANBL and XG1 MM cells (kindly provided by John Shaughnessy, University of Arkansas for Medical Sciences, Little Rock, AR) in myeloma medium consisting of RPMI 1640, 10% FBS, 10% MarrowMax supplement, 100 μM pyruvate, 100 units/ml penicillin G, and 100 μg/ml streptomycin (Invitrogen). In some cases, we added 1 ng/ml IL-6 (Sigma, St. Louis, http://www.sigmaaldrich.com) to the expansion medium.
Production of Dkk1
We stably transfected COS cells (Fugene 6; Roche, Indianapolis, http://www.roche.com) with a construct encoding human Dkk1 in pcDNA3.1 (Invitrogen) and grew them in large-scale cell factories (Cell Factory; Nunc, Roskilde, Denmark, http://www.nuncbrand.com) in α-MEM containing 10% (vol/vol) FBS, 100 μg/ml streptomycin, 100 units/ml penicillin G, 2 mM glutamine, and 800 μg/ml G418 (Sigma). At confluency, we transferred the cells to 1 l of serum-free medium and allowed them to secrete Dkk1 for 48 hours. We then dialyzed the Dkk1-containing medium into 10 mM ammonium carbonate (pH 8.0) (Sigma) using a customized large-scale dialysis apparatus and concentrated it by partial freeze-drying (ThermoSavant ModulyoD; Thermo Electron Corporation, Waltham, MA, http://www.thermo.com). We assayed the concentrated sample by Dkk1 enzyme-linked immunosorbent assay (ELISA) . Purity averaged 80%, and yield averaged 4 mg/liter of medium. We produced protein preparations from untransfected COS cells in parallel with the Dkk1 extracts to serve as controls and found that control preparations contained no Dkk1 and had no effect in the mineralization assay (data not shown). We freeze-dried aliquots of Dkk1 at 10 μg per tube and prepared control medium using equivalent volumes.
We plated cells at 5,000 cells per cm2 in 10-cm2 six-well plates and allowed them to adhere to the plastic. The next day, we replaced the medium. Two to three days later, the cultures attained approximately 70% confluence, and we started the assay. On day 1, we replaced the medium with mineralization medium consisting of CCM supplemented with 50 μg/ml l-ascorbic acid and 5 mM β-glycerophosphate (Sigma). Some cultures received Dkk1, LiCl (Sigma), or BIO (Calbiochem, San Diego, CA, http://www.emdbiosciences.com). We then replaced the medium on days 3, 5, 7, and 9. On day 10, we washed the monolayers first with 10 ml phosphate-buffered saline (PBS) and then with 5 ml wash buffer consisting of 100 mM Tris base (pH 10), 100 mM NaCl, and 1 mM MgCl2 (Sigma). We quantified alkaline phosphatase activity by adding 1 ml paranitrophenol phosphate in diethanolamine buffer (One Step; Pierce, Rockford, IL, http://www.piercenet.com) directly onto the monolayer. We monitored light absorbance at 405 nm by spectrophotometry (FLUOstar; BMG Labtech, Durham, NC, http://www.bmglabtech.com) and determined reaction rates as the initial rate of change of absorbance of 405 nm light per minute. We normalized reaction rates to cell numbers, which we determined by using fluorescent dye intercalation (CyQuant; Molecular Probes, Eugene, OR, http://probes.invitrogen.com) to assay for DNA content. We related cell numbers to DNA content by measuring the fluorescence of a set of samples of known cell number ranging from 10,000 to 625 cells and compared fluorescence of experimental samples to the fluorescence per cell curve to determine the cell number of experimental samples. The mineralized monolayers secreted vast amounts of matrix and required overnight digestion at 37°C with 200 μg/ml proteinase K (Invitrogen) in a digestion buffer composed of 100 mM NaCl, 100 mM Tris [pH 8], 10 μM CaCl2, and 0.005% Triton X-100 (Sigma) prior to fluorescence incorporation assay (CyQuant). Due to the large variation in matrix secretion between donors and experiments, we normalized the assay to cell number rather than to total protein. To validate the measurement of alkaline phosphatase as an osteogenic marker, we exposed confluent cultures to CCM or osteogenic medium for a further 14 days. Using a slight modification of a previous protocol , we then stained the cultures with Alizarin Red for calcified matrix and back-extracted the stain. Next, we stained the destained monolayers for total protein using Crystal Violet and back-extracted the dye. We calculated the ratio of optical densities of neutralized Alizarin Red to Crystal Violet to get a measure of the degree of mineralization in the cultures and expressed the data as the mean of three assays. The error bars represent the maximum variation possible after ratio calculation.
Growth Curves and Production of MSC-Conditioned Medium
We prepared MSC-conditioned medium (MSC-CM) by collecting CCM exposed for 2 days to confluent cultures of MSCs. We diluted the CCM to 10% FBS with serum-free CCM, then diafiltered at 4°C using a tangential-flow filtration system fitted with 150-cm2 polyvinylidene difluoride 5-kDa filters (Millipore, Billerica, MA, http://www.millipore.com) against 10 changes of myeloma medium without FBS. We then filtered the medium through a 0.22-μm filter. For proliferation experiments, we plated MM cells at 150 cells per cm2 in six-well plates with medium as indicated in Results. For each condition, we harvested cells every 24 hours and then washed and froze the cells until the end of the experiment, at which time we determined cell numbers by DNA fluorescence incorporation (CyQuant). For IL-6-blocking experiments, we harvested the cells after 4 days. We used IL-6-blocking antibody (cat. no. I2143; Sigma) at 10 μg/ml for IL-6-blocking experiments.
We recovered about 5 × 104 MM cells by centrifugation, counted them using a hemacytometer, washed them in PBS, and then suspended them in Laemmli sample buffer (Invitrogen) with complete protease inhibitor (Roche). We separated the samples on 4%–12% Bis-Tris polyacrylamide using the 4-morpholineethanesulfonic acid buffer system (Novex; Invitrogen). For detection of Dkk1, IL-6, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH), we used R&D AF1096 (R&D Systems, Minneapolis, http://www.rndsystems.com), Sigma no. I2143, and Chemicon clone 6C5 (Chemicon, Temecula, CA, http://www.chemicon.com), respectively, all at 1:1,000 dilution. We used peroxidase-conjugated secondary antibodies at 1:2,500 dilution and detected the signal as described previously  (Sigma). We normalized load volumes of media to cell number.
We carried out conventional reverse transcription-polymerase chain reaction (RT-PCR) assays for IL-6  and GAPDH  as previously described. We carried out real-time RT-PCR on the ABI Prism 7700 Sequence Detector using the following Assays-on-Demand primer sets: IL-6, Hs00174131_m1; DKK1, Hs00183740_m1. We normalized all expression levels to GAPDH using either Hs99999905_m1 or Endogenous Control no. 4310884E. All reactions used manufacturer-recommended universal cycling parameters. Briefly, we incubated samples for 30 minutes at 60°C to allow reverse transcription to take place, then ran 40 cycles at 94°C for 20 seconds and at 62°C for 1 minute. We defined the critical threshold as 10 standard deviations above the baseline fluorescence of cycles three to 15.
We measured the level of IL-6 protein in cell culture supernates by ELISA using a kit (Quantikine; R&D Systems). Osteoprotegerin (OPG) levels were measured by ELISA using MAB8051 as the capture antibody and BAF805 as the detection antibody (R&D Systems) in a sandwich ELISA carried out in surface-treated 96-well plates (Immunomodule; Nunc). Streptavidin-conjugated horseradish peroxidase (R&D Systems and Pierce) was used to bind the biotinylated detection antibody and convert 2,2′-azino-di(3-ethylbenzthiazoline-6-sulfonate) substrate (1-Step ABTS; Pierce). We derived a quantitative relationship between protein levels and the change in absorbance per minute at 405 nm using the FLUOstar plate reader (BMG Labtech). Recombinant OPG reference material was acquired from R&D Systems.
We labeled MM cells with CellTracker Green (Molecular Probes) and incubated them in an environmentally controlled chamber. We recorded images with a charge-coupled-device camera (ORCA ER; Hamamatsu, Bridgewater, NJ, http://www.sales.hamamatsu.com) and controlled microscopic functions with software (METAMORPH; Universal Imaging, Sunnyvale, CA, http://www.moleculardevices.com). For immunocytochemical labeling of β-catenin, we cultured MSCs for 4 days on chamber slides, then treated as indicated for 15 hours. We then removed the medium and washed the cells in PBS (Invitrogen) prior to fixation with 4% paraformaldehyde (USB, Cleveland, http://www.usbweb.com) for 12 minutes. The cells were then permeabilized and blocked with PBS containing 0.5% Triton X-100 (Fisher Scientific, Pittsburgh, http://www.fishersci.com) and 0.5% (vol/vol) goat serum (Chemicon). β-Catenin was detected with a Cy3-conjugated β-catenin mouse monoclonal antibody (Clone 15B8; Sigma) for 90 minutes at a 1:200 dilution in PBS. We washed the slides three times with 1 ml PBS after each step and mounted the slides in vectashield mounting medium with 4′-6-diamidino-2-phenylindole (DAPI; Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com). We captured images using a fluorescent microscope (Nikon Eclipse E800; Melville, NY, http://www.nikon.com) with Spot-RT imaging software (Diagnostic Instruments, Inc., Sterling Heights, MI, http://www.diaginc.com), using identical gain and exposure times. We prepared composite figures using Adobe Photoshop (Adobe Systems Incorporated, San Jose, CA, http://www.adobe.com) and then processed the composite image to remove background and improve contrast. The image from the red channel only is presented for clarity; however, the DAPI signal overlaid the nuclear β-catenin signal in all cases.
In initial experiments, we confirmed that MM cell lines express Dkk1. Two MM cell lines, XG1 and ANBL1, expressed high levels of both the message and the protein (Fig. 1). We then examined the effect of conditioned medium from undifferentiated MSCs (MSC-CM) on expression of Dkk1 by MM cells. We recovered the MSC-CM after a 2-day exposure to confluent MSCs and processed it as detailed in Materials and Methods. Interestingly, MM cells produced more Dkk1 in the presence of MSC-CM (Fig. 1).
We then designed an in vitro assay to determine the consequences of exposing MSCs to high levels of Dkk1. In these experiments, we treated semiconfluent MSCs with osteogenic medium and either vehicle, Dkk1, inhibitors of GSK3-β, or Dkk1 together with inhibitors of GSK3-β. In earlier experiments (data not shown), we found that an increase in membrane alkaline phosphatase activity (Fig. 2A), which peaked at approximately day 10, preceded calcium deposition (Fig. 2B), which reached maximal levels at approximately day 24. We therefore used alkaline phosphatase activity as an early marker of osteogenic differentiation of MSCs. We found that high levels of Dkk1 prevented the increase in alkaline phosphatase activity that occurred in control osteogenic cultures. As expected, BIO, a potent and selective inhibitor of GSK3-β that has been used to activate Wnt signaling in embryonic stem cells , was able to restore alkaline phosphatase activity to normal levels. However, LiCl, widely used to inhibit GSK3-β and activate Wnt signaling [17–20], did not restore alkaline phosphatase activity. In fact, levels were reduced (Fig. 2A). Immunocytochemical labeling of MSCs confirmed that BIO treatment resulted in β-catenin stabilization and that Dkk1 treatment resulted in degradation of β-catenin, at the doses used in our assay (Fig 2C). Other inhibitors tested either were less effective or were toxic to MSCs (data not shown).
We next examined the consequences of extended exposure of MM cells to MSC-CM to determine whether MSCs produce any factors that would stimulate the proliferation of MM cell lines in vitro. We grew cultures of MM cells under four conditions: in RPMI with 10% FBS and 100 μM pyruvate (MM medium), in MM medium with 1 ng/ml IL-6, in MSC-CM diafiltered (> 5-kDa fraction) into MM medium, or in MSC-CM diafiltered into MM medium and diluted 1:5 with MM medium. We harvested cells every 24 hours and determined cell numbers by a DNA fluorescence incorporation assay. We found that the > 5-kDa protein component of MSC-CM had a strong stimulatory effect on the proliferation of MM cells (Fig. 3A).
Because IL-6 is known to be an important growth and survival factor for MM cells [7, 11, 21, 22], we hypothesized that IL-6 would be the likely candidate for the growth-promoting component of MSC-CM. We cultured MM cells for 4 days in MSC-CM in the presence or absence of an IL-6-blocking antibody and found that the IL-6-blocking antibody inhibited the MSC-CM-enhanced proliferation of MM cells (Fig. 3B), whereas a control antibody had no effect on proliferation. To confirm the production of IL-6 by MSCs, we harvested MSCs, extracted RNA, and determined IL-6 expression levels by RT-PCR and Western blotting. We confirmed that undifferentiated MSCs expressed high levels of IL-6 message (Fig. 3C).
Fully mature osteoblasts produce only small amounts of IL-6 . In agreement with this observation, we found that levels of IL-6 message were reduced in MSCs differentiated with osteogenic medium (Fig. 4A) but remained high if Dkk1 was added to the osteogenic medium (Fig. 4A). Addition of BIO decreased IL-6 expression to below control levels, apparently by releasing MSCs from the MM proliferation-promoting cycle induced by Dkk1. To examine this observation in more detail, we tested the effects of BIO on very early cultures of differentiating MSCs in which IL-6 is transiently upregulated prior to terminal differentiation and arrest of IL-6 production. Measurement of IL-6 protein in supernatants from cultures of differentiating MSCs confirms that they initially secrete high levels of IL-6, but secrete much less IL-6 if BIO is present in the medium, consistent with an earlier initiation of the osteogenic differentiation program (Fig. 4B). Interestingly, BIO is a member of the aryl hydrocarbon family of molecules, which have been shown to suppress IL-6 expression via the aryl hydrocarbon receptor , but the contribution of this effect to the suppression of IL-6 has not been determined in MSCs. Given that canonical Wnt signaling regulates osteoclast development by stimulating production of OPG by osteoblasts , we also measured secreted OPG in the same cell culture supernatants. In contrast to the effect of BIO on IL-6 production, we found that the concentration of OPG in the MSC-CM gradually increased upon culture of MSCs with osteogenic medium and was significantly elevated with BIO treatment (Fig. 4B).
Tian et al.  first reported the high level of expression of the Wnt inhibitor Dkk1 in MM cells and suggested that it may affect osteogenesis. We confirmed the secretion of Dkk1 in two MM cell lines and found that the presence of MSC-CM enhanced the production of Dkk1. Using an in vitro osteogenesis assay, we showed that treating MSCs with Dkk1 inhibited the osteogenic differentiation of the cells. BIO, a potent and selective inhibitor of GSK3-β, blocked the effects of Dkk1. Surprisingly, LiCl did not block the action of Dkk1 but resulted in a further inhibition of differentiation. In preliminary experiments, we also found that administration of 20 mM LiCl in drinking water did not stimulate osteogenesis in a transgenic mouse model of osteogenesis imperfecta (Emigdio Reyes and Carl Gregory, Tulane University, New Orleans, unpublished data).
We found that MSC-CM promoted the proliferation of two MM cell lines in vitro and that this stimulation of growth was principally due to IL-6 production by MSCs. Previous reports described a positive feedback loop between bone marrow stromal cells similar to MSCs and MM cells leading to enhanced stromal expression of IL-6 [6, 7, 11, 12, 25]. The inhibition of osteogenesis demonstrated here, in conjunction with higher osteoclast activity  and vascular endothelial growth factor expression , may promote proliferation, vascularization, and metastasis.
When MM cell lines are co-cultured with a monolayer of MSCs, the MM cells adhere to the MSCs and proliferate, becoming a component of the monolayer (Fig. 5A). Therefore, the MM cells take full advantage of the high local concentrations of the soluble factors secreted by MSCs, and the MSCs are, in turn, exposed to the maximal dose of Dkk1 from the MM cells, resulting in a cyclic interaction that may explain the aggressive and irreversible destruction of the bone in MM (Fig. 5B) [12, 25, 26]. Furthermore, maintenance of MSCs in an immature state by exposure to high levels of Dkk1 may predispose the cells respond in a pro-osteoclastogenic fashion , further decreasing the osteoblast-to-osteoclast ratio near the site of osteolytic lesions. Current treatment strategies for MM involve inhibition of osteoclast activation and minimizing disease progression by lowering available IL-6 levels. However, these strategies have no effect on repair of existing osteolytic lesions [10, 28]. Use of a Dkk1-blocking agent, such as BIO, holds promise both for enhancing current therapies that only slow disease progression and for enabling future therapies that may permit repair of existing osteolytic lesions.
The work was supported in part by NIH Grants P40 RR 17447 and R01 AR48323, Hospital Corporation of America, and the Louisiana Gene Therapy Research Consortium.
The authors indicate no potential conflicts of interest.