Dkk-1-derived Synthetic Peptides and Lithium Chloride for theControl and Recovery of Adult Stem Cells from Bone Marrow*

Received for publication, June 4, 2004, and in revised form, October 20, 2004Published, JBC Papers in Press, October 25, 2004, DOI 10.1074/jbc.M406275200

Carl A. Gregory, Anthony S. Perry, Emigdio Reyes, Adam Conley, W. Grady Gunn,and Darwin J. Prockop‡

From the Center for Gene Therapy, Tulane University Health Sciences Center, New Orleans, Louisiana 70130

It is established that human mesenchymal stem cells(hMSCs) from bone marrow are a source of osteoblastprogenitors in vivo and under appropriate conditions,differentiate into osteoblasts ex vivo. Because hMSCsare recovered by iliac crest aspirate and enriched byvirtue of their adherence to tissue culture plastic, thecells provide a convenient ex vivo model for the study ofosteogenic tissue repair in an experimentally accessiblesystem. Recent advances in the field of skeletal develop-ment and osteogenesis have demonstrated that signal-ing through the canonical wingless (Wnt) pathway iscritical for the differentiation of progenitor cell linesinto osteoblasts. Inhibition of such signals can predis-pose MSCs to cell cycle entry and inhibit osteogenesis.Here, we report that synthetic peptides derived from thesecond cysteine-rich domain of the canonical Wnt inhib-itor Dickkopf-1 (Dkk-1) have utility in controlling thegrowth and recovery of hMSCs from bone marrowstroma. Three peptides corresponding to residues 217–269 in Dkk-1 were each found to enhance the prolifera-tion of hMSCs in culture over 2 days. The most activepeptide exhibited agonistic characteristics in that it ab-lated the proliferation lag observed when cultures ofhMSCs receive fresh medium. It also reduced the ex-pression of endogenous Dkk-1 (Gregory, C. A., Singh, H.,and Prockop, D. J. (2003) J. Biol. Chem. 278, 28067–28078). When the cytosolic level of �-catenin was ele-vated by addition of LiCl to cultures of hMSCs, the pep-tide also accelerated degradation of �-catenin onwithdrawal of lithium. A second peptide, correspondingto residues 184–204 had preferential and high affinityfor hMSCs in the log phase of proliferation. Peptideoverlay assays on hMSC lysates confirmed that the pep-tide bound to a 184-kDa protein corresponding to themolecular mass of LRP6. Cells recovered by this peptidehad enhanced osteogenic potential but less chondro-genic potential compared with controls. Because Wntantagonists increase the number of non-committedhMSCs in culture, they may be of use in increasing therate of osseous wound healing in vivo by increasing thelevel of systemically migrating hMSCs. Therefore, suchmolecules could contribute to the development of anovel family of pharmaceutical agents for the improve-ment of the healing process in humans.

Mesenchymal stem cells or marrow stromal cells (humanMSCs)1 can differentiate into numerous mesenchymal tissuelineages including osteoblasts, chondrocytes, adipocytes, andneural precursors (1–8) making them attractive candidates forcytotherapy, bioengineering, and gene therapy (9). In previousstudies on secreted factors that regulate the growth of hMSCs,we found that synthesis of the canonical Wnt inhibitor Dick-kopf-1 (Dkk-1) was required before the cells entered the cellcycle (10). Based on this observation, we investigated the role ofcanonical Wnt signaling in hMSC growth and differentiationinto osteoblasts. In addition to its proliferative properties, ca-nonical Wnt signaling has been implicated in a variety of de-velopmental processes (11). More specifically, Wnt signalingapparently plays an essential role in the differentiation ofpluripotent cell lines into osteoblasts (12, 13) and natural in-hibitors of Wnt signaling can inhibit osteogenesis (13, 14).Indeed, it has been shown that a high systemic level of Dkk-1is likely to contribute to the persistence of osteolytic lesions incases of multiple myeloma (14). Conversely, mutations in theLRP6 receptor that prevent the binding of the canonical Wntinhibitor cause abnormally high bone density (15). CanonicalWnt signaling is activated by binding of some forms of the Wntligands to the receptors, Frizzled and LRP6, that in turn re-cruit the cytoplasmic bridging molecule, Dishevelled. In theactive form, Dishevelled inhibits the action of glycogen synthe-tase kinase 3, which in turn, reduces the phosphorylation of�-catenin (16, 17). Because only phosphorylated �-catenin isdegraded by the ubiquitin-mediated pathway (18, 19), the lev-els of unphosphorylated �-catenin in the cytosol are raised.There are two known fates for cytosolic �-catenin: one is thatthe protein may be localized to the membrane where it contrib-utes to the formation of adherens junctions that mediate inter-cellular contacts and may play a role in transducing externalsignals into the cell (20). The other fate is that stabilized�-catenin translocates to the nucleus where it complexes withtranscription factors that either promote differentiation or pro-liferation (21, 22).

Dkk-1 apparently acts to prevent the coalescence of LRP6and Frizzled by binding to LRP6 via its carboxyl-terminalcysteine-rich (Cys2) domain (23, 24). To determine whetherpeptide derivatives of Dkk-1 can agonize or antagonize Dkk-1,we synthesized seven biotinylated 21-mer peptides correspond-ing to the Cys2 domain of the protein. The peptides were

* This work was supported by the American Stem Cell TherapyFoundation, National Institutes of Health Grants AR47796 and AR48323, the Oberkotter Foundation, HCA, the Health Care Company,and the Louisiana Gene Therapy Research Consortium. The costs ofpublication of this article were defrayed in part by the payment of pagecharges. This article must therefore be hereby marked “advertisement”in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

‡ To whom correspondence should be addressed. E-mail: dprocko@tulane.edu.

1 The abbreviations used are: MSC, marrow stromal cells; hMSC,human marrow stromal cells; Dkk-1, Dickkopf-1; LRP6. low densitylipoprotein-related protein 6; Cys2, carboxyl-terminal cysteine-rich do-main; MALDI-TOF, matrix-assisted laser desorption ionization time offlight; PBS, phosphate-buffered saline; BSA, bovine serum albumin;bis-Tris, 2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)propane-1,3-diol; FACS, fluorescence-activated cell sorter; MES, 4-morpho-lineethanesulfonic acid; GAPDH, glyceraldehyde-3-phosphate dehydro-genase; RT, reverse transcriptase; ALP, alkaline phosphatase.

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 280, No. 3, Issue of January 21, pp. 2309–2323, 2005© 2005 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

This paper is available on line at http://www.jbc.org 2309

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

assessed for hMSC bioactivity in ex vivo assays of hMSC bind-ing, osteogenic differentiation, �-catenin localization, and pro-liferation. Three peptides corresponding to residues 217–269 inDkk-1 were each found to enhance the proliferation of hMSCsin culture over 2 days. The most active peptide exhibited ago-nistic characteristics in that it ablated the proliferation lagobserved when cultures of hMSCs receive fresh medium. It alsoreduced the expression of endogenous Dkk-1 (10). When thecytosolic level of �-catenin was elevated by addition of LiCl tocultures of hMSCs, the peptide also accelerated degradation of�-catenin on withdrawal of lithium. A second peptide, corre-sponding to residues 184–204, had a preferential and highaffinity for hMSCs in the log phase of proliferation. Peptideoverlay assays on hMSC lysates confirmed that the peptidebound to a 184-kDa protein corresponding to the molecularmass of LRP6. Magnetic activated cell sorting using the peptideas an affinity ligand allowed the enrichment of hMSCs fromhuman bone marrow aspirates with �50% efficiency. To testwhether positive canonical Wnt signaling affected differentia-tion of hMSCs to osteoblasts, we mimicked the effect of Wnt bythe addition of LiCl to cultures of differentiating hMSCs. In thepresence of LiCl and osteo-inductive medium, the expression ofalkaline phosphatase by hMSCs was activated and reachedmaximal expression in 1 week compared with 3 weeks in con-trols. The results suggest the use of agents that regulate thecanonical Wnt pathway for control and recovery of hMSCs.Peptides corresponding to Dkk-1, other Wnt inhibitors, or theWnt ligands themselves could contribute to the development ofa novel class of pharmaceutically active compounds for thecontrol of tissue repair in humans. Furthermore, previouslydescribed agents that affect the canonical Wnt pathway maybe useful for controlling hMSCs in the treatment of bone frac-tures and skeletal lesions such as those observed in multiplemyeloma (14).

MATERIALS AND METHODS

Tissue Culture of hMSCs—Bone marrow aspirates of �2 ml weredrawn from healthy donors ranging in age from 19 to 49 years under anInstitutional Review Board approved protocol. The adherent nucleatedcells were separated from the aspirate and cultured in �-minimal es-sential medium containing 20% (v/v) fetal calf serum as describedpreviously (8, 25).

After 14 days in culture, adherent cells were recovered by incubationwith 0.25% (w/v) trypsin and 1 mM EDTA (Fisher) for 5 to 7 min at 37 °C(Fisher). For the analysis of hMSCs at the log phase of growth, the cellswere re-plated in 54-cm2 plates (Costar, Fisher) at a density of 100 cellsper cm2 and allowed to adhere for 24 h. The medium was then replacedwith fresh medium containing various concentrations (4–15 mM) oftissue culture grade LiCl or KCl (Fisher) and allowed to proliferate for7 days prior to recovery and cell counting or Western blotting. Forosteogenesis assays, the cells were plated at a density of 5000 cells percm2 and allowed to form a confluent monolayer for 14 days with changesof medium every 3–4 days. For Alizarin Red S staining, the cultureswere established in 6-well plates (Costar, Fisher) and for RNA extrac-tion, the cultures were established in 54-cm2 plates (Costar).

Peptides—Seven peptides (21 residues long) were synthesized byTufts University Medical School Core Facility (Boston, MA) using anABI 431 Peptide synthesizer employing FastMoc chemistry. To facili-tate the synthesis of some of the peptides, some cysteine residues weresubstituted by serine residues. The peptides were biotinylated at theamino terminus and purified by reverse phase high performance liquidchromatography. To confirm purity and identity, the peptides weresubjected to matrix-assisted laser desorption ionization time of flight(MALDI-TOF) mass spectrometry (Ciphergen Chip Reader, CiphergenBiosystems, Freemont, CA) as described previously (10).

Immunocytochemistry—hMSCs were seeded at 1000 cells per cm2 in4-cm2 chamber slides (Nunc, Fisher). When the hMSCs established amonolayer of the appropriate density, the slides were washed in PBS andthe cells fixed in phosphate-buffered 4% (v/v) formaldehyde for 15 min.The cells were washed, then incubated for 1 h in a block buffer consistingof PBS containing 0.01% (v/v) Triton X-100 (Sigma) and 5% (v/v) normalgoat serum (Chemicon International, Temecula, CA). After block, the

slides were incubated in anti-�-catenin monoclonal antibody at a dilutionof 1 in 800 (clone 5H10, Chemicon) for 2 h. The slides were then washed3 times in block buffer and incubated in Alexa Fluor 594- (red) or 486(green)-conjugated anti-mouse secondary antibody (Molecular Probes). Insome cases, rhodamine-conjugated phalloidin (Molecular Probes) wasused to stain actin filaments at a concentration of 1:1000 in PBS. Slideswere dried and mounted with 4�,6-diaminido-2-phenylindole containingmounting reagent (Vector Laboratories, Burlingame, CA) and visualizedon an upright epifluorescent microscope (Eclipse 800, Nikon). For peptidebinding assays, the hMSCs were grown to a density of �3000 cells per cm2

(log phase) or 20,000 cells per cm2 in 10-cm2 6-well plates (Costar, Fisher).The monolayers were fixed in phosphate-buffered 4% (v/v) formaldehydefor 2 min only. The monolayers were blocked in peptide block bufferconsisting of PBS containing 5% (w/v) highly purified BSA (Sigma) for 1 h.Solutions (10 �g ml�1) of biotinylated peptides were prepared in peptideblock buffer and incubated with the hMSCs for 1 h. The slides werewashed 3 times, and then incubated in Alexa Fluor 594-conjugatedstreptavidin (Molecular Probes).

Peptide Overlay Assays—A lysate of hMSCs was prepared from cul-tures at the log phase of growth. Approximately 3 million hMSCs on six154-cm2 tissue culture dishes were washed in PBS and scraped into 10ml of PBS. The cells were recovered by centrifugation at 500 � g for 15min and resuspended in 1 ml of PBS containing 2% (v/v) Triton X-100,1% (w/v) SDS, 1 mM EDTA (pH 8.0), and the equivalent of 1 half-tabletof Complete protease inhibitor (Roche). The lysate was sheared bypasses through a 18-gauge needle and incubated at 4 °C for 30 min priorto electrophoresis and blotting. Fifteen �l (45,000 cells) were loadedonto a 4–12% NuPage bis-Tris polyacrylamide gel (Invitrogen, Carls-bad, CA) and blotted onto polyvinylidene difluoride. The filter waswashed in PBST (PBS containing 0.1% (v/v) Tween 20), blocked inPBST containing 5% (w/v) powdered milk, then incubated in PBSTcontaining 5% (v/v) normal goat serum (Chemicon) for 15 h at 4 °C torecover some of the structure of the denatured proteins. Each peptidewas incubated with one of the blotted lanes at 10 �g ml�1 in PBST for2 h at room temperature. The blots were washed three times in PBS,and then incubated in PBST containing streptavidin-conjugated horse-radish peroxidase at a dilution of 1:2000 (Pierce) for 1 h at roomtemperature. After three washes in PBST, the blots were incubated inperoxidase substrate (100 mM Tris, pH 8.0, containing 75 �M paracou-maric acid, 500 �M luminol, and 0.006% (v/v) hydrogen peroxide, Sigma)for 5 min prior to exposure to photographic film (Kodak autoradiogra-phy, Sigma) or digital imaging (Typhoon Imager, Bio-Rad).

Magnetic Cell Sorting—Bone marrow aspirates (about 2 ml) wererecovered from the iliac crest. The nucleated cells were enriched bydensity gradient centrifugation as described previously (8, 25), washedin incubation buffer of PBS containing 1 mM EDTA and 0.5% (w/v)highly purified BSA, and recovered by centrifugation at 500 � g for 15min. Up to 2 million cells were incubated with incubation buffer con-taining 10 �g ml�1 peptide B for 1 h. After washing in excess washbuffer (PBS containing 1 mM EDTA) the cells were separated by anautomated magnetic-activated cell sorter fitted with a column contain-ing streptavidin-conjugated beads (Miltenyi Biotech, Auburn, CA). ThePOSSELS program was used with wash buffer as the mobile phase.Negative controls lacked peptide but were separated using identicalprotocols. Conditioned medium was prepared by incubating standardfresh hMSC medium with hMSCs at the log phase of growth for 2 days.After 2 days, the medium was filtered through a 0.2-�m2 membraneand flash frozen with liquid nitrogen prior to use.

Fluorescence-activated Cell Sorting—FACS sorting was achieved us-ing an automated instrument (Cytomics FC 500, Beckman Coulter).The phycoerythrin-conjugated ABCG2 transporter antibody (clone 5D3)was purchased from eBioScience (San Diego, CA). All other fluorophore-conjugated antibodies used for FACS analysis, including isotype con-trols, were purchased from BD Pharmingen (San Diego, CA). Approxi-mately 50,000 hMSCs were incubated in the presence of themanufacturer’s recommended concentration of antibody diluted in PBScontaining 1% (w/v) BSA and 1 mM EDTA for 1 h at room temperature.Prior to FACS analysis, the cells were washed in excess PBS containing1% (w/v) BSA and 1 mM EDTA.

Colony Forming Unit Assay—Colony forming unit assays were car-ried out as described previously (10). Briefly, hMSCs were counted byhemacytometer and 100 cells were transferred to a 154-cm2 tissueculture dish (Costar, Fisher) containing 30 ml of culture medium. After3 weeks, the dishes were washed in PBS and 1% (v/v) crystal violet in50% methanol was added to stain the colonies. After 15 min, the disheswere washed thoroughly with water and allowed to dry. Colonies over2-mm diameter were counted.

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells2310

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

Osteogenic Differentiation and Quantification of Alizarin Red SStaining—For standard osteogenic differentiation, confluent monolay-ers of hMSCs were incubated in medium supplemented with 10�8 M

dexamethasone, 50 �g ml�1 ascorbic acid, and 5 mM �-glycerol phos-phate (Sigma) for 21 days with changes of medium every 5 days. Fortesting the effects of lithium, cultures were incubated under the sameconditions but with the addition of 10 mM LiCl or KCl and incubation forup to 30 days. Quantification of staining was carried out using amodified version of a previously described procedure (27). Briefly, thecellular aggregates were washed in PBS, pressed flat with a Teflon-coated spatula, and fixed in formalin for 15 min. The cells were thenstained with 40 mM Alizarin Red S for 30 min and washed 4 times withPBS. The stained cells were then transferred to a 2-ml screw-top mi-crocentrifuge tube and incubated at 85 °C for 15 min in 1 ml of 10% (v/v)acetic acid overlaid with 0.5 ml of light mineral oil. The extraction wascooled on ice and then centrifuged at 21,000 � g and 0.5 ml of thesupernatant was transferred to a fresh tube containing 10% (v/v) 10%ammonium hydroxide. The red solution was then transferred to a96-well plate and read at 450 nm on a plate reader (Bio-Rad).

Adipogenic Differentiation and Quantification of Oil Red O Stain-ing—All reagents were purchased from Sigma. Confluent monolayers ofhMSCs in 6-well plates (10 cm2 per well) were incubated in mediumsupplemented with 10�8 M dexamethasone, 5 � 10�8 M isobutylmeth-ylxanthine, and 5 � 10�7 M indomethacin. After 21 days, the adipogeniccultures were fixed in 10% formalin for 15 min and stained with freshOil Red-O solution in 60% (v/v) isopropyl alcohol in PBS for 20 min. Thedishes were washed 3 times with excess PBS. Stained dishes were

extracted with 2 ml of extraction buffer containing 50% (v/v) ethanoland 2% (w/v) SDS for 15 min at room temperature. Three aliquots (200�l) of the extracted Oil Red O was transferred to a 96-well plate andquantified by absorbance measurement at 405 nm (Bio-Rad). The linearrange of detection was determined by standard solutions of Oil Red O.

Chondrogenic Differentiation—Chondrogenic differentiation wascarried out in accordance with Sekiya et al. (8) on 200,000 cells recov-ered by either peptide B recovery or standard means.

Evaluation of Cell Number—For assay of proliferation, cells werequantified by fluorescent labeling of nucleic acids (CyQuant dye; Mo-lecular Probes) using a described previously method (10) and a micro-plate fluorescence reader (FLX800; Bio-Tek Instruments Inc., Winooski,VT) set to 480 nm excitation and 520 nm emission. Data from 3 separateassays were statistically analyzed using the two-tailed Student’s t test.Experiments were repeated using hMSCs from two donors.

Extraction of Cytoskeletal Fractions and Western Blotting—Triton-insoluble fractions were prepared using a protocol by Ko et al. (26) withmodifications (10). Briefly, one million cells were suspended in 1 ml ofice-cold PBS with protease inhibitors (Roche Diagnostics) and 1% (v/v)Triton X-100 (Sigma). Lysis proceeded for 10 min on ice followed by a60-s centrifugation at 800 � g to remove particulate bodies. The cy-toskeletal pellet was separated from the cytoplasmic fraction by cen-trifugation at 14,000 � g for 15 min and resuspended in PBS containing0.1% SDS. The concentration of protein was measured by Bradfordassay (Sigma) prior to Western blotting. Electrophoresis was carriedout using commercial reagents and systems (Novex; Invitrogen). Ap-proximately 10 �g of protein were added to the appropriate volume of

FIG. 1. Design of Dkk-1-derivedpeptides. Panel a, diagram of the Dkk-1molecule illustrating the presence of twocysteine-rich domains separated by a“hinge” region. The LRP6 binding domainis in the second cysteine-rich (Cys2) do-main. Panel b, the sequence of the secondcysteine-rich domain of Dkk-1 (bold) withthe sequences of seven overlapping pep-tides corresponding to the Dkk-1 Cys2 se-quence. A list of the peptides synthesizedand tested in this study is given below thesequence. Serines substituted from cys-teines to facilitate synthesis are denotedin lowercase type. Panel c, MALDI-TOFmass spectrometry of the synthesizedpeptides demonstrate that the calculatedmolecular masses correspond to the ob-served masses to within 2 Da.

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells 2311

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

FIG. 2. Binding of hMSCs by the Dkk-1-derived peptides. Panel a, whole cell binding assay of peptide binding to intact hMSCs in monolayerculture at the log phase of growth. The biotinylated peptides were incubated with the cells, then binding was detected by fluorescently conjugatedstreptavidin. Fluorescent signal (left) is presented next to the phase image (right). The control (ctrl) was carried out in the absence of peptide. Panelb, peptide overlay assay demonstrating that peptide B and to a lesser extent, peptide E recognizes a 184-kDa protein on blotted whole cell lysatesof hMSCs at the log phase of growth. The molecular mass of LRP6 is �185 kDa. Equal loading of the blots was confirmed by Ponceau Red staining

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells2312

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

4� SDS-PAGE sample buffer containing 1 mM 2-mercaptoethanol(Sigma). The samples were heated at 100 °C for 2 min and electrophore-sed on a 4–12% NuPage bis-Tris gel using the MES buffering systemfollowed by transfer to polyvinylidene difluoride for 1 h. Filters wereblocked in PBST containing 5% (w/v) powdered milk (Bio-Rad) for 1 h.For detection of �-catenin, blots were probed with an anti-�-cateninmonoclonal antibody at a dilution of 1 in 1000 (clone 5H10, Chemicon).For detection of actin, blots were probed with a monoclonal antibody ata dilution of 1 in 1000 (clone AC-10, Sigma). For detection of GAPDH,blots were probed with a monoclonal antibody at a dilution of 1 in 1000(clone 6C5, Chemicon International). For secondary detection, a rabbitanti-mouse IgG antibody coupled to horseradish peroxidase was used ata dilution of 1:1500 (Sigma). After three washes in PBST, the blots wereincubated in peroxidase substrate for 5 min prior to exposure to photo-graphic film (Kodak autoradiography, Sigma) or digital imaging (Ty-phoon Imager, Bio-Rad).

RT-PCR Analysis—Total mRNA was prepared from 1 million cellsusing a commercially available kit (High Pure; Roche Diagnostics). Thelysates were briefly centrifuged at 15,000 � g to remove excess matrixprotein prior to RNA purification. A one-tube RT-PCR reaction (Titan;Roche Diagnostics) was employed for the synthesis of cDNA and PCRamplification. Reactions were carried out on a thermal cycler (AppliedBiosystems 9700; PE Applied Biosystems, Foster City, CA). For ampli-fication of GAPDH, the following primers were used: ccccttcattgacct-caact (sense) and cgaccgtaacgggagttgct (antisense) with the followingconditions: initial cDNA synthesis, 50 °C for 45 min; denaturation 95 °Cfor 1 min; annealing 52 °C for 1 min; and extension 72 °C for 1 min for28 cycles. For amplification of collagen X, the following primers wereused: ggtaacaggaatgcctgtgtctg (sense) and tcacattggagccactaggaa (anti-sense) using the following conditions: initial cDNA synthesis, 50 °C for45 min; denaturation 95 °C for 1 min; annealing 55 °C for 1 min; andextension 72 °C for 1 min for 35 cycles. For amplification of osteonectin,the primer sequences and conditions of Bilbe et al. (28) were used; foralkaline phosphatase, the primers and conditions of Fromigue et al. (29)were used; and for connexins 43 and 45, the primers and conditions ofOviedo-orta et al. (30) were used. For amplification of Dkk-1 the primersand conditions of Gregory et al. (10) were used.

Histochemistry—Micromass pellets were fixed in phosphate-bufferedformaldehyde (10% v/v) overnight and set in liquid paraffin. Sections (10�m) were prepared using a microtome (Microm International, Walldorf,Germany) and placed on microscope slides (Fisher). The slides were pro-cessed for staining as follows. The slides were de-paraffinized in Clear-Rite (Microm) and gradually hydrated to 95% (v/v) ethanol. Slides wererinsed in dH2O, and stained with 1% (v/v) Alizarin Red S (pH 4.1),Toluidine Blue, or Safranin O (Sigma) for 10 min. After washing in dH2O,the sections were counterstained in 1% (v/v) light green solution (Sigma)for 1 min, washed in dH2O, and dehydrated to 100% ethanol. The sectionswere finally cleared with Clear-Rite and coverslipped in Permount mount-ing medium (ProSciTech, Queensland, Australia).

RESULTS

Peptide Design—Because Dkk-1 acts to prevent the coales-cence of LRP6 and Frizzled by binding to LRP6 via its carboxyl-terminal cysteine-rich (Cys2) domain (23, 24) (Fig. 1, panel a),7 overlapping 21-mer peptides corresponding to the Cys2 do-main of the ligand were synthesized (Fig. 1, panel b). Thepeptides were conjugated to biotin at the amino terminus andin some cases, cysteines were substituted by serine to facilitatesynthesis and purification by reverse phase high performanceliquid chromatography. The peptides were designated A to Gfrom the amino to carboxyl terminus of the Cys2 domain (Fig.1, panel c). The highly purified peptides were subjected toMALDI-TOF mass spectrometry to confirm identity and purity.In each case the molecular mass was within 2 daltons of thecalculated mass (Fig. 1, panel c) and the purity was greaterthan 99.5%.

Binding of hMSCs by the Synthetic Peptides—To assess theaffinity of the synthetic peptides for hMSCs, log phase cultures

were prepared in 6-well plates and fixed. Monolayers wereincubated with the peptides (10 �g ml�1) and then washedprior to incubation in fluorophore-conjugated streptavidin. Oninspection of the stained monolayers by epifluorescent micros-copy, binding by peptides B and E was evident (Fig. 2, panel a).A peptide overlay assay on log phase hMSC lysates separatedby SDS-PAGE and blotted onto polyvinylidene difluoride, indi-cated that a protein with the molecular mass of LRP6 (184kDa) had affinity for peptide B and to a lesser extent, peptideE (Fig. 2, panel b). Peptide B had a greater affinity for rapidlydividing hMSCs than those in the stationary phase of growth,distinctly staining 100% of all log phase hMSCs (semiconflu-ent) but only a very small fraction (about 2%) of hMSCs onconfluent monolayers (Fig. 2, panel c).

Separation of Highly Clonogenic hMSCs from the NucleatedFraction of Bone Marrow Based on Affinity to Peptide B—Rap-idly dividing hMSCs in culture are highly clonogenic and effi-ciently differentiate into osteoblasts (25). Therefore, becausepeptide B had preferential affinity for rapidly dividing hMSCs,it is possible that peptide B may be useful for the recovery of asubset of such cells directly from bone marrow extracts. Thenucleated fraction of whole bone marrow was prepared fromaspirates by discontinuous density gradient centrifugation anddiluted in PBS containing 1 mM EDTA and 0.5% (w/v) highlypurified BSA in the presence or absence of 10 �g ml�1 peptideB. After incubation for 1 h, the cells were washed and subjectedto magnetic-activated cell sorter using an automated instru-ment loaded with paramagnetic beads coated with streptavi-din. A scheme for the recovery of hMSCs from bone marrowextracts is presented in Fig. 3 (panel a). The binding fractionwas seeded into 45-cm2 tissue culture dishes and incubated for2 days in the presence of 50% fresh hMSC tissue culture me-dium supplemented with 50% hMSC preconditioned tissue cul-ture medium prepared as described under “Materials andMethods.” After 2 days, clusters of adherent cells could bevisualized by phase-contrast microscopy (Fig. 3a, left frame)with a small number of non-adherent cells floating in themedium. At day 2, the medium was replaced by fresh 50%conditioned medium and incubated for a further 2 days prior toreplacement with fresh unconditioned medium. Thereafter, themedium was replaced every 3 days for 2 weeks resulting inrapidly dividing adherent cells within colonies that rapidlycoalesced to establish a partial monolayer (Fig. 3a, rightframe). Negative control extractions that lacked peptide did notestablish rich cultures, but there were a few adherent cellspresent that did not readily divide.

Data from FACS analysis of the adherent cells from 2-weekcultures demonstrated a cell surface phenotype consistent withhMSCs. The cells expressed CD90 and HLA I but not CD34,CD45, HLA II, CD109, CD117 (Kit), or the Hoescht 33342effluxing ABC transporter, suggesting that the cells are MSCsand not directly related to hematopoietic tissue (34, 35) (Fig. 3,panel b). To measure the frequency of hMSCs with affinity forpeptide B, hMSCs bound to the peptide were extracted from 2million nucleated bone marrow cells, allowed to adhere to tis-sue culture plastic for 15 h, and then counted by the fluores-cence incorporation assay. Routinely, �200 cells were recov-ered in these experiments. To test the efficiency of theextraction, 2 million bone marrow cells were spiked with10,000 rapidly dividing hMSCs that were recovered by scrap-

and GAPDH detection (below). Panel c, assay of peptide B binding to intact hMSCs at log phase (log) and stationary (confluent) phase of growth(conf). The biotinylated peptides were incubated with the cells then binding was detected by fluorescently conjugated streptavidin. Fluorescentsignal (red) from peptide B is presented with 4�,6-diaminido-2-phenylindole signal (blue) to visualize nuclei. The control was carried out in theabsence of peptide. Note that all of the log phase (semiconfluent) hMSCs bind the peptide readily, whereas only a small number of cells have welldefined affinity for peptide B in stationary (confluent) monolayers of hMSCs. The experiments were repeated on cells from two separate donors.

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells 2313

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

ing, then subjected to peptide B mediated recovery. Approxi-mately 5,000 cells were recovered in each instance, suggestingthat the procedure had an efficiency of about 50%. Correctingfor an efficiency of 50%, the frequency of hMSCs in the nucle-ated fraction of bone marrow with affinity for peptide B wastherefore �2 in 10,000 (Fig. 4, panel a). Recovery of hMSCs bythe standard method of plastic adherence for 15 h in culturewas also assayed by the same method. The percentage recoveryranged from 0.1 to 0.2%, roughly 5–10 times the recovery withpeptide B. Although the plates were washed thoroughly withPBS to remove non-adherent cells, the MSCs frequently bind tohematopoietic cells when adhering to plastic and this phenom-enon could contribute to the elevated fraction of cells detectedby the original assay. Furthermore, larger, less clonogenic cellspresent in the aspirate may be recovered by plastic adherencebut not by peptide B affinity, also resulting in higher initialrecoveries by the original method.

The hMSCs Recovered by Peptide B Affinity Differentiate intoAdipocytes and Osteoblasts—To determine whether the hMSCs

recovered by peptide B affinity were multipotent, they wereinduced to differentiate into mesenchymal tissue lineages. Un-der standard conditions, the hMSCs differentiated into adiposetissue, as assayed by Oil Red O staining and osteoblasts, asassayed by Alizarin Red S staining and chondrocytes, as as-sayed by Toluidine Blue and Safranin O (Fig. 5, panels a–c).When compared with MSCs recovered from the same donor bystandard methods, it was found that peptide B recovered cellsmore readily differentiated to osteoblasts with about 2-foldhigher efficiency (Fig. 5, panel c) but this was at the expense ofchondrogenic potential (Fig. 5, panel b). Indeed, peptide Brecovered MSCs only partially differentiated into chondrocytescapable of depositing sulfated proteoglycan (purple with Tolu-idine Blue and Red with Safranin O) suggesting that the pep-tide B selection process enriched for cells with a propensity toform bone but not cartilage.

Peptides Mapping to the Carboxyl-terminal End of the Dkk-1Cys2 Domain Mimics the Activity of Dkk-1 in hMSCs—BecauseDkk-1 was previously shown to predispose hMSCs to entry into

FIG. 3. Recovery of hMSCs from whole bone marrow aspirates by peptide B-mediated magnetic cell sorting. Panel a, strategy forrecovery of hMSCs from whole bone marrow using magnetic activated cell sorting (above). The nucleated cells from bone marrow aspirates areincubated with biotinylated peptide B then recovered by paramagnetic beads coated with streptavidin. Phase-contrast micrographs of therecovered hMSCs after 48 h incubation (lower left) and 2 weeks (lower right) are presented. Panel b, fluorescence activated cell sorting of hMSCsafter 2 weeks in culture. The cells express CD90 and HLA I but not CD34, CD45, HLA II, CD109, CD117 (Kit), or the Hoescht 33342 effluxing ABCtransporter suggesting that the cells are MSCs and not directly related to hematopoietic tissue (34, 35). The appropriate isotype control profile ispresented with each assay. The experiments were repeated on cells from two separate donors.

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells2314

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

the cell cycle, the peptides were tested in hMSC proliferationassays (10). When hMSCs were plated at 500 cells per cm2 inmedium containing 2% (v/v) fetal bovine serum, and 10 �g ml�1

of peptides corresponding to the carboxyl-terminal portion of

the Cys2 domain (from Cys217–Ala269, peptides D–F), prolifer-ation was accelerated resulting in a 30% increase (p � 0.01) incell number after 2 days (Fig. 6, panel a). The increase in hMSCrecovery was probably attributable to an ablation of the 12-h

FIG. 4. hMSCs recovered from whole bone marrow aspirates based on their affinity for peptide B are more clonogenic than hMSCsrecovered by conventional means. Panel a, recovery of hMSCs with affinity for peptide B from 2 million whole bone marrow cells. The fractionof cells recovered from the whole bone marrow extract is �0.02%. Results on cells from two separate donors are presented and data are means fromthree replicate experiments. Plastic adherent cells were counted after 15 h in culture. Panel b, clonogenic potential of hMSCs recovered usingpeptide B compared with hMSCs recovered by the standard method of plastic adherence. The experimental conditions used to carry out theexperiment are presented as a flow diagram (left) and the colony forming unit assays are presented (right) for hMSCs from 2 donors recovered bythe standard method and based on affinity to peptide B. p values: p � 0.05 (*), p � 0.01 (**) for n � 5.

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells 2315

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

FIG. 5. Differentiation of peptide B recovered hMSCs into osteoblasts and adipocytes. Panel a, cultures of peptide B recovered andstandard hMSCs were expanded to the stationary phase of growth and incubated in adipocytic medium for 21 days. The monolayers were thenstained with Oil Red O to detect fat droplets. Micrographs (right) demonstrate lipid formation by peptide B-derived cells in the presence, but not

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells2316

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

lag phase of growth observed on changes of medium whenDkk-1 levels are suboptimal (10). A similar lag phase ablationcould be observed when hMSCs from two donors were exposedto medium containing 10 �g ml�1 peptide E, whereas controlcultures treated with peptide A exhibited the lag phase (Fig. 6,panel b). Furthermore, endogenous Dkk-1 transcription wasdown-regulated in response to peptide E treatment as assayedby RT-PCR (Fig. 6, panel c). This is consistent with the obser-vation that Dkk-1 expression is regulated by negative feedback(10). Also peptide E induced an overall reduction of �-catenin inhMSCs when assayed by immunocytochemistry similar to theeffect seen with recombinant Dkk-1 (10).

The Effect of LiCl and Dkk-1-derived Peptides on hMSCs inCulture—In previous studies, BMP2 has been shown to induceosteogenic differentiation in a canonical Wnt-dependentmanner in a number of mesenchymal progenitor cell lines (12,13). Furthermore, inhibition of the canonical Wnt signalingpathway by Dkk-1 was shown to inhibit such differentiation(12–14) and predispose hMSCs to cell cycle entry (10). To inves-tigate the effect of canonical Wnt signaling on hMSC differen-tiation and to evaluate the effects of Dkk-1-derived peptidesfurther, Wnt signaling was mimicked by the addition of LiCl torapidly dividing cultures of hMSCs (31). In initial assays ofproliferation, 10–15 mM lithium reduced the rate of hMSCdivision in a dose-dependent manner over 7 days, whereascontrol cultures treated with KCl were unaffected (Fig. 7, panela). Concentrations above 20 mM caused complete cell deathover 7 days (data not shown). To test whether LiCl treatmenthad an effect on the levels of cytosolic �-catenin, a fractionationtechnique was employed based on detergent extraction andWestern blot assays of the cytoskeletal and cytosolic proteins(Fig. 7, panel b) (26). The fractionation was evaluated by sim-ultaneous Western blot assays of actin and GAPDH. GAPDHwas not detected in the cytoskeletal fraction where the majorityof actin was detected. The cytosolic fraction contained all of thedetectable GAPDH and some actin, presumably not associatedwith microfilaments. In the KCl-treated control cultures, themajority of the �-catenin was present in the insoluble fraction,possibly as actin-associated adherens junctions, and trace lev-els were detectable in the soluble fraction of the cells (Fig. 7,panel b). LiCl-treated cells also had high levels of �-cateninassociated with the insoluble fraction but there were also highsoluble levels. The observation that cytosolic �-catenin wasincreased in response to lithium is consistent with the conclu-sion that activation of canonical Wnt signaling inhibits hMSCproliferation. Because positive canonical Wnt signaling seemsto enhance osteogenic differentiation of a number of mesenc-hymal cell lines (12–13), the effect of lithium on hMSCs ininhibiting proliferation probably represents an initial stage ofosteogenic differentiation.

The effect of peptide E on �-catenin distribution was evalu-ated in hMSCs pretreated with LiCl. After pretreatment ofhMSCs with lithium to increase cytosolic �-catenin, the hMSCswere transferred to fresh medium containing 10 �g ml�1 ofpeptide E or peptide A as a control. After 6 h, the hMSCs wererecovered and assayed for �-catenin distribution. The level ofsoluble �-catenin was reduced in peptide E-treated hMSCs,

suggesting that its degradation had been accelerated (Fig. 7,panel c). This observation is consistent with the prediction thatpeptide E inhibits canonical Wnt signaling by preventing theWnt-mediated formation of the Frizzled/LRP co-receptor com-plex and thus reducing the activity of Dishevelled in inhibitingglycogen-synthetase kinase 3�.

The Effect of LiCl on Osteogenesis in an Osteogenic Differen-tiation Assay—To investigate the effect of lithium directly onosteogenesis, the hMSCs were grown to confluence and treatedfor up to 30 days with an osteogenic medium containing a100-fold lower than standard concentration of dexamethasoneand 10 mM LiCl or 10 mM KCl. Lower levels of dexamethasonewere used to improve the detection of differences in osteogen-esis induced by LiCl. Under these conditions, the hMSC mono-layer detached from the plastic and spontaneously curled intoa roughly spherical cellular aggregate. In the presence of LiCl,the aggregate was formed after about 7 days of treatment,whereas in the presence of KCl, the effect was seen after about12 days. The aggregates were fixed, paraffin embedded, sec-tioned, and stained for calcified deposits (Alizarin Red S), sul-fated proteoglycan (Toluidine Blue), and collagenous deposits(Trichrome). On sectioning and staining, the overall morphol-ogy of the LiCl-treated aggregates were surprisingly differentfrom the controls in that they were much more compact (Fig. 8,panel a), suggesting that the intercellular contacts were muchmore adherent. Control micromasses had a more open, lamel-lar, morphology (Fig. 8, panel a). Staining of the sections withAlizarin Red S revealed that mineralization was much moreevident in the LiCl-treated aggregates (Fig. 8, panel a) than inthe control with dense patches of mineral detected throughoutthe sections of the LiCl-treated cells. In both cases, there waslittle evidence of sulfated proteoglycan deposition as demon-strated by Toluidine Blue (data not shown). The matrix adja-cent to the constituent cells in the LiCl-treated aggregatesappeared to be collagenous in nature as demonstrated byTrichrome staining (data not shown).

To semiquantify the apparent enhancement of Alizarinstaining in LiCl-treated cultures of hMSCs, Alizarin Red S wasextracted and measured over a time course of osteogenic dif-ferentiation. Over 27 days, the LiCl-treated cultures exhibitedan enhanced rate of mineralization when compared with KCl-treated cells as determined by Alizarin Red S staining (Fig. 8,panel b). Untreated hMSCs in the absence of osteogenic sup-plements also exhibited a degree of mineralization but to alesser extent than the cultures incubated in osteogenic medium(Fig. 8, panel b), demonstrating that osteogenesis is the defaultfate of hMSCs grown for long periods in stationary cultures. Toinvestigate the effect of LiCl treatment more closely, total RNAwas extracted from the mineralizing micromasses at 10-dayintervals. The transcription of genes known to play a role inosteogenic development was assayed by RT-PCR. Because ofthe surprising enhancement of cellular adhesion, the transcrip-tion of gap junctional proteins, connexin 43 and connexin 45were assayed, because both of these proteins are known to playa role in osteogenesis (32). The levels of connexin 43 remainedconstant throughout the time course of osteogenesis. Interest-ingly, connexin 45 was present in untreated cultures but was

absence, of adipocytic medium. Back extraction of the dye at weekly intervals and colorimetric quantification (left) demonstrate extensive Oil RedO staining throughout both the monolayers (n � 6, error bars � 1 S.D.) Duplicate experiments were carried out on two separate donors. Panel b,pellets (200,000 cells) of peptide B recovered and standard hMSCs were incubated for 21 days in chondrocyte medium. The pellets were thenstained with Toluidine Blue and Safranin O to evaluate proteoglycan deposition. Panel c, cultures of peptide B recovered and standard hMSCs wereexpanded to the stationary phase of growth and incubated in osteogenic medium for 21 days. The monolayers were then stained with Alizarin RedS to detect mineral. Micrographs (right) demonstrate mineral deposition by peptide B-derived cells in the presence, but not absence, of osteogenicmedium. Back extraction of the dye at weekly intervals and colorimetric quantification (left) demonstrate extensive Alizarin Red S stainingthroughout the monolayer (n � 6, error bars � 1 S.D.) that was more intense in the peptide B recovered cells than the control. p values: p � 0.05(*), p � 0.01 (**) for n � 6. Duplicate experiments were carried out on cells from two separate donors.

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells 2317

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

FIG. 6. Peptide E mimics the action of Dkk-1 on hMSCs. Panel a, fluorescence incorporation assay demonstrating that overlapping peptidescorresponding to amino acids 217–269 (peptides D, E, and F) increase the rate of hMSC proliferation. Log phase cultures of hMSCs were incubatedfor 2 days in medium containing 2% (v/v) fetal calf serum in the presence of vehicle or 10 �g ml�1 of each peptide. The cells were then recoveredand counted. p values versus control: p � 0.05 (*), p � 0.01 (**) for n � 3. Experiments were duplicated with cells from two donors. Panel b,fluorescence incorporation assay demonstrating that peptide E ablates the lag phase of hMSC growth observed on replacement of medium. Logphase cultures of hMSCs were incubated for 24 h in fresh medium containing 2% (v/v) fetal calf serum in the presence of vehicle or 10 �g ml�1 ofpeptide E or peptide A (control). After 12 and 24 h, hMSCs were recovered and counted. Data on cells from two donors are expressed as the meanof 3 counts with error bars representing 1 S.D. Statistically significant differences in data are p � 0.05 (*). Panel c, RT-PCR assay demonstratingthat peptide E reduced the rate of endogenous Dkk-1 expression. Log phase cultures of hMSCs were incubated for 8 h in fresh medium containing2% (v/v) fetal calf serum in the presence of vehicle or 10 �g ml�1 of peptide E or peptide A (control). Data on cells from 2 donors are presented induplicate with GAPDH levels to control for mRNA quality. Panel d, immunohistochemistry of late log phase cultures of hMSCs demonstrates thatpeptide E reduces the level of cytoplasmic �-catenin. �-Catenin is stained green, actin is stained red, and the nuclei are stained blue.

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells2318

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

completely down-regulated on exposure to osteogenic medium(Fig. 8, panel c). Surprisingly, LiCl failed to effect transcriptionof either connexin 43 or connexin 45 and therefore did notappear to account for the enhanced cellular adherence or min-

eralization on treatment with LiCl. RT-PCR assays for os-teonectin demonstrated that lithium treatment reduced therate of its transcription at all of the time intervals tested (Fig.8, panel c). Because lithium has roughly the same effect as the

FIG. 7. Peptide E accelerates the degradation of cytosolic �-catenin induced by lithium pretreatment. Panel a, proliferation assaydemonstrating that lithium inhibits the proliferation of log phase hMSCs in a dose-dependent manner. The hMSCs were plated at 5000 cells percm2 and incubated for 4 days in the presence of various concentrations of LiCl or KCl (control) prior to counting of cells by fluorescenceincorporation assay. p values versus appropriate KCl control: p � 0.05 (*), p � 0.005 (***) for n � 6. Experiments were duplicated with 2 donors.Panel b, Western blot assay of whole (w), soluble (s), and insoluble (i) fractions of hMSCs incubated with 10–15 mM LiCl or KCl (control)demonstrating that lithium treatment increases the soluble level of �-catenin. The efficiency of extraction is evaluated by probing the same blotsimultaneously for actin and GAPDH. Panel c, Western blot assays of whole, soluble, and insoluble preparations of hMSCs demonstrating that 10�g ml�1 peptide E, but not peptide A (control), accelerates the degradation of cytosolic �-catenin induced by lithium pretreatment. The efficiencyof extraction is evaluated by probing the same blot simultaneously for actin and GAPDH. The experimental outline is presented above the blots.Experiments were duplicated with cells from 2 donors. TX100, Triton X-100.

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells 2319

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

FIG. 8. Long-term treatment of hMSCs with 10 mM LiCl improves the differentiation of hMSCs into osteoblasts. Panel a, ARS sectionsof mineralizing micromass cultures treated with 10 mM LiCl or KCl (control). Note that lithium treatment results in a more tightly packed andintensely stained culture. Panel b, extraction and colorimetric quantification of ARS confirms that lithium increases the rate of osteogenicdifferentiation by hMSCs. p values versus appropriate “Supplemented medium” control: p � 0.05 (*), p � 0.01 (**) for n � 6. Experiments wereduplicated with 2 donors. Supplemented medium refers to complete medium containing osteogenic supplements as defined under “Materials andMethods.” Panel c, RT-PCR assays for ALP demonstrates that addition of 10 mM lithium causes earlier and higher levels of ALP expression inmineralizing cultures of hMSCs. Data are presented with cells from 2 separate donors. Conn43 and Conn45, connexins 43 and 45.

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells2320

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

EF-hand domain of osteonectin on the development of Xenopuslaevis embryos (33), it is possible that lithium treatment re-sults in negative feedback of the expression of osteonectin. Thetranscription of type X collagen was up-regulated in osteogenicmedium but unaffected by the addition of lithium, whereasalkaline phosphatase (ALP) activity was profoundly up-regu-lated. In KCl-treated cultures, ALP transcription was maxi-mally up-regulated after 30 days of mineralization, whereas inthe presence of lithium, the level of ALP transcription reachedmaximal transcription by day 10 (Fig. 8, panel c). These obser-vations are in agreement with previous work demonstratingthe requirement for sustained levels of cytosolic and nuclear�-catenin for ALP transcription (12).

The Effect of Peptide E on hMSC Osteogenesis—Initial ex-periments on stationary cultures of hMSCs in mineralizingconditions failed to detect an effect of peptide E on osteogenicdifferentiation (data not shown). Because peptide E was previ-ously shown to act on rapidly dividing hMSCs (Fig. 6, panel b),log phase cultures were assayed. LRP6 expression is highest inrapidly dividing hMSCs and therefore the effect of peptide E inagonizing the action of Dkk-1 on LRP6 should be more readilydetected. Cells were plated at 1000 cells per cm2 in 6-wellplates with complete medium and allowed to adhere for 15 h.The next day, medium was replaced by osteogenic mediumcontaining 50 �g ml�1 peptide E or peptide A as a control. At7-day intervals, the hMSCs were recovered from 3 wells of a6-well plate and assayed for cell number. The cells in theremaining 3 wells were fixed and stained with Alizarin Red Sfor dye extraction and quantification of mineralization. Thelevel of Alizarin Red S staining per cell was calculated andplotted for 21 days. In this modified assay, the hMSCs re-mained as a monolayer and mineralized with far less efficiencythan stationary phase hMSCs but mineralization could still bedetected by Alizarin Red S staining after 7 days of treatment(Fig. 9, inset). On quantifying the Alizarin Red staining, pep-tide E reduced the rate of osteogenic differentiation by hMSCswhen compared with peptide A (Fig. 9), providing further evi-dence that peptide E inhibits the effect of canonical Wnt sig-naling during osteogenesis.

DISCUSSION

Because hMSCs can differentiate into numerous mesenchy-mal tissue lineages including osteoblasts, chondrocytes, adipo-

cytes, and neural precursors (1–8), there has been much inter-est in the cells and their potential application in cytotherapy,bioengineering, and gene therapy (9). The rapid expansion ofhMSCs under simple culture conditions is an attractive char-acteristic of the cells and it has been the subject of muchinvestigation (1, 8, 10). In investigating secreted factors thatregulate the growth of hMSCs, it was found that synthesis ofthe canonical Wnt inhibitor Dkk-1 was required before the cellsentered the cell cycle (10). Furthermore, Wnt signaling appar-ently plays an essential role in the differentiation of pluripo-tent cell lines into osteoblasts (12–14). Indeed, it has beenshown that a high systemic level of Dkk-1 contributes to thepersistence of osteolytic lesions in cases of multiple myeloma(14), and mutations in the LRP6 receptor that prevents thebinding of Dkk-1 cause abnormally high bone density (15).

In this study, we sought to design overlapping peptides de-rived from the Cys2 domain of Dkk-1 and characterize theireffects on hMSC proliferation and differentiation. In brief, fourpeptides of interest were identified. One peptide (Leu184–Ser204, peptide B) had a high specific affinity for the receptorLRP6 and rapidly dividing hMSCs. Furthermore, it was effec-tive as an affinity molecule for the recovery of highly clonogenichMSCs from whole bone marrow aspirates. A further threepeptides, corresponding to the extreme carboxyl terminus(Cys217–Arg237, peptide D; Cys233–Cys253, peptide E; andGlu249–Ala269, peptide F), increased the rate of proliferation ofhMSCs. In ex vivo bioassays, peptide E was found to robustlyexhibit bioactivity consistent with it acting as a Dkk-1 agonist.

Peptide B was initially identified by virtue of its high affinityfor fixed intact hMSCs in monolayer culture (Fig. 2, panel a). Inthe same assay, the peptide was found to be specific for semi-confluent, rapidly dividing hMSCs and although the peptidehad affinity for a few isolated cells in stationary phase cultures,the peptide did not bind the majority of cells in the confluentmonolayer (Fig. 2, panel c). Furthermore, peptide B bound to a184-kDa species when used to probe blotted lysates of log phasehMSCs (Fig. 2, panel b), suggesting that the peptide had affin-ity for its intended target, the 184-kDa LRP6 receptor. Thesedata are in agreement with the observation that LRP6 expres-sion is high in rapidly dividing hMSC cultures and low instationary phase cultures (10). The detection of a few isolatedhMSCs with affinity for peptide B on stationary phase mono-

FIG. 9. Treatment of log phase hMSCs with peptide E inhibits the differentiation of hMSCs into osteoblasts. Extraction andcolorimetric quantification of Alizarin Red S demonstrates that peptide E decreases the rate of osteogenic differentiation by dividing hMSCs. Dataare expressed as the mean of 3 Alizarin Red S readings divided by the mean of three CyQuant cell counts. Error bars represent the maximum errorpossible based on the 2 S.D. Statistically significant differences in data are p � 0.01 (**). Phase micrographs are presented (inset) demonstratingthe presence of clusters of mineralizing cells at day 14 in the control that are not present on treatment with peptide E.

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells 2321

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

layers demonstrates variable but generally low levels of expres-sion of LRP6 under these conditions. The peptide was used torecover hMSCs with high clonogenic potential from the nucle-ated fraction of bone marrow. MSCs recovered by this methodconstituted about 0.02% of the total number of cells in the bonemarrow extract, �5–10-fold lower than the traditional recoverytechnique involving only adherence to plastic. The higher re-covery when solely adherence is used is probably because of thepresence of hMSCs with a lower clonogenic potential thatwould not be selected by peptide B. Furthermore, it is likelythat hematopoietic cells are carried through with the hMSCs inthe initial stages of plating, also accounting for the increasedinitial cell recovery. Peptide B-recovered cells had a higherclonogenic potential than standard preparations of hMSCs(Fig. 4, panel b), confirming that the peptide had excludedhMSCs with lower clonogenic potential. The highly clonogenichMSCs were also pluripotent in that they readily differentiatedinto bone, adipose, and chondrocytes (Fig. 5). When comparedwith control MSCs recovered by standard means, the peptideB-recovered cells more efficiently deposited a calcium-rich ma-trix during osteogenic assays in vitro, but they less readilydifferentiated into chondrocytes capable of depositing sulfatedproteoglycan. Because the selection process enriches for cellsexpressing the LRP6 receptor, a known regulator of osteogen-esis in humans (15, 16), it is likely that the peptide enriches foran osteogenic progenitor subpopulation with a propensity toform bone but not cartilage. Because peptide B recovers hMSCswith a significantly higher clonogenic potential, it is likely tohave future utility in hMSC isolation and also measurement ofhMSC levels in bone marrow and blood.

In assays of hMSC proliferation a further three peptides ex-hibited bioactivity consistent with a Dkk-1 agonist. The peptidescorresponding to the extreme carboxyl terminus (Cys217– Arg237,peptide D; Cys233–Cys253, peptide E; and Glu249– Ala269, peptideF), increase the initial rate of proliferation of hMSCs in ex vivobioassays (Fig. 6, panel a). The action of the peptides was shortlived and could be detected only for up to 2 days in time courses,presumably because of the short half-life in the medium. PeptideE also behaved in a similar manner to Dkk-1 by ablating the 12-hlag phase in proliferation detected on a change of culture medium(Fig. 6, panel b) (10) and by reducing the endogenous level ofDkk-1 transcription. Based on these observations it is probablethat peptide E mimics the activity of Dkk-1 by preventing thebinding of Wnt ligands or the co-receptor Frizzled to the targetmolecule, LRP6. To explore the action of peptide E further, itseffect on �-catenin levels and distribution was examined. Be-cause inhibition of canonical Wnt signaling has a profound effecton �-catenin levels, a Dkk-1-like inhibitor of Wnt signaling wouldtherefore act to release the inhibition of glycogen synthetasekinase 3 and accelerate �-catenin phosphorylation and degrada-tion. Addition of lithium to hMSCs mimicked positive Wnt sig-naling by increasing the level of unphosphorylated �-catenin inthe cytosol through inhibition of glycogen synthetase kinase 3(Fig. 7, panel b). Furthermore, on withdrawal of the lithium andaddition of peptide E, the lithium-induced accumulation of cyto-solic �-catenin was degraded more rapidly than in culturestreated with a control peptide. These data strongly suggest thatpeptide E acts as a Wnt inhibitor in a manner similar to Dkk-1itself.

In previous studies, BMP2 has been shown to induce osteo-genic differentiation in a canonical Wnt-dependent manner in anumber of mesenchymal progenitor cell lines (12, 13). Further-more, inhibition of the canonical Wnt signaling pathway byDkk-1 was shown to inhibit such differentiation (12–14) andpredispose hMSCs to cell cycle entry (10). Consistent withthese observations, we observed that addition of lithium to cells

at the log phase of growth reduced the rate of proliferation in adose-dependent manner (Fig. 7, panel a). Incubation of conflu-ent cultures in the presence of standard osteogenic mediumwith 100-fold lower than standard concentration of dexametha-sone (1 � 10�10 M) resulted in detachment of the monolayerfrom the plastic and formation of a roughly spherical aggregateof hMSCs that formed a calcified matrix as determined byAlizarin Red S staining. Incubation of hMSCs under the sameconditions but with 10–15 mM LiCl formed a similar, but muchmore compact ball of mineralizing cells. Quantification of min-eral by colorimetric measurement of Alizarin Red S stainingdemonstrated that LiCl-treated cultures produced mineralizedmatrix more rapidly than the controls. Analysis of gene expres-sion by the differentiating hMSCs revealed that transcriptscommonly associated with osteogenesis increased over time inosteogenic medium both in the presence and absence of LiClbut there was a striking up-regulation of alkaline phosphatasetranscription. Alkaline phosphatase transcription was maxi-mal after 10 days in lithium-treated culture compared with 30days in the controls, an observation consistent with cell linestudies (12, 13).

Because inhibition of canonical Wnt signaling has beenshown to inhibit the differentiation of osteogenic precursors toosteoblasts in a number of systems (12–14), we therefore de-cided to test the effects of peptide E on osteogenic differentia-tion of hMSCs. In initial experiments on stationary cultures ofhMSCs, peptide E had no effect on osteogenic differentiation(data not shown). The reason for this is unclear, but it is likelythat in the process of becoming a stationary phase culture,hMSCs irreversibly commit to an osteogenic lineage. Indeed,long-term cultures of hMSCs spontaneously form mineral evenin the absence of osteo-inductive supplements (27). It is likelythat inhibitors of Wnt signaling work to sustain the undiffer-entiated state of hMSCs before an irreversible cascade of os-teogenic commitment, therefore log phase cultures were as-sayed. Although hMSCs did not efficiently form mineralizeddeposits in proliferative conditions, peptide E treatment re-sulted in even lower levels of mineralization suggesting thatthe action of peptide E was confined to non-committed anddividing hMSCs.

In this study, we describe the production of synthetic inhib-itors of the canonical Wnt signaling pathway based on peptideanalogues of Dkk-1. The agents described here that affect thecanonical Wnt pathway may be useful for controlling hMSCs inthe treatment of bone fractures and skeletal lesions such asthose observed in multiple myeloma. They also may have util-ity in producing subpopulations of MSCs with a greater pro-pensity for osteogenesis.

REFERENCES

1. Friedenstein, A. J., Gorskaga, U., and Kalungina, N. N. (1976) Exp. Hematol.4, 267–274

2. Ashton, B. A., Allen, T. D., Howlett, C. R., Eaglesom, C. C., Hattori, A., andOwen, M. (1980) Clin. Orthop. 151, 294–307

3. Bennet, J. H., Joyner, C. J., Triffit, J. T., and Owen, M. E. (1991) J. Cell Sci.99, 131–139

4. Piersma, A. H., Brockbank, K. G., Ploemacher, R. E., van Liet, E., Brakel-vanPeer, K. M., and Visser, P. J. (1985) Exp. Hematol. 13, 237–243

5. Caplan, A. I. (1991) J. Orthop. Res. 9, 641–6506. Houghton, A., Oyajobi, B. O., Foster, G. A., Russell, R. G., and Stringer, B. M.

(1998) Bone 22, 7–167. Kopen, G. C., Prockop, D. J., and Phinney, D. G. (1999) Proc. Natl. Acad. Sci.

U. S. A. 96, 10711–107168. Sekiya, I., Vuoristo, J. T., Larson, B. L., and Prockop, D. J. (2002) Proc. Natl.

Acad. Sci. U. S. A. 99, 4397–44029. Prockop, D. J. (1997) Science 276, 263–272

10. Gregory, C. A., Singh, H., and Prockop, D. J. (2003) J. Biol. Chem. 278,28067–28078

11. Cadigan, K. M., and Nusse, R. (1997) Gene Dev. 11, 3286–330512. Bain, G., Muller, T., Wang, X., and Papkoff, J. (2003) Biochem. Biophys. Res.

Commun. 301, 84–9113. Rawadi, G., Vayssiere, B., Dunn, F., Baron, R., and Roman-Roman, S. (2003)

J. Bone Miner. Res. 18, 1842–185314. Tian, E., Zhan, F., Walker, R., Rasmussen, E., Ma, Y., Barlogie, B., and

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells2322

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

Shaughnessy, J. (2003) N. Engl. J. Med. 349, 2483–249415. Boyden, L. M., Mao, J., Belsky, J., Mitzner, L., Farhi, A., Mitnick, M. A., Wu,

D., Insogna, K., and Lifton, R. P. (2002) N. Engl. J. Med. 346, 1513–152116. Stambolic, V., Ruel, L., and Woodgett, J. R. (1996) Curr. Biol. 6, 1664–1668;

Correction (1997) Curr. Biol. 7, 19617. Klein, P. S., and Melton, D. A. (1996) Proc. Natl. Acad. Sci. U. S. A. 93,

8455–845918. Huelsken, J., and Birchmeier, W. (2001) Curr. Opin. Genet. Dev. 11, 547–55319. Kikuchi, A. (2000) Biochem. Biophys. Res. Commun. 268, 243–24820. Nagafuchi, A. (2001) Curr. Opin. Cell Biol. 13, 600–60321. Van Noort, M., and Clevers, H. (2002) Dev. Biol. 244, 1–822. Pandur, P., Maurus, D., and Kuhl, M. (2002) Bioessays 24, 881–88423. Brott, B. K., and Sokol, S. Y. (2002) Mol. Cell. Biol. 22, 6100–611024. Li, L., Mao, J., Sun, L., Liu, W., and Wu, D. (2002) J. Biol. Chem. 277,

5977–598125. DiGirolamo, C. M., Stokes, D., Colter, D., Phinney, D. G., Class, R., and

Prockop, D. J. (1999) Br. J. Haematol. 107, 275–281

26. Ko, K., Arora, P., Lee, W., and McCulloch, C. (2000) Am. J. Physiol. 279,C147–C157

27. Gregory, C. A., Gunn, W. G., Peister, A., and Prockop, D. J. (2004) Anal.Biochem. 329, 77–84

28. Bilbe, G., Roberts, E., Birch M., and Evans, D. B. (1996) Bone 19, 437–44529. Fromigue, O., Marie, P. J., and Lomri, A. (1997) Cytokine 9, 613–62330. Oviedo-Orta, E., Hoy, T., and Evans, W. H. (2000) Immunology 99, 578–59031. Jope, R. S. (2003) Trends Pharmacol. Sci. 9, 441–44332. Lecanda, F., Towler, D. A., Ziambaras, K., Cheng, S., Koval, M., Steinberg,

T. H., and Civitelli, R. (1998) Mol. Biol. Cell 9, 2249–225833. Huynh, M. H., Sage, E. H., and Ringuette, M. (1999) Dev. Growth Differ. 41,

407–41834. Goodell, M. A., Brose, K., Paradis, G., Conner, A. S., and Mulligan, R. C. (1996)

J. Exp. Med. 183, 1797–180635. Murray, L. J., Bruno, E., Uchida, N., Hoffman, R., Nayar, R., Yeo, E. L., Schuh,

A. C., and Sutherland, D. R. (1999) Exp. Hematol. 27, 1282–1294

Peptide Agonists of Dkk-1 in Mesenchymal Stem Cells 2323

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from

Darwin J. ProckopCarl A. Gregory, Anthony S. Perry, Emigdio Reyes, Adam Conley, W. Grady Gunn and

Recovery of Adult Stem Cells from Bone MarrowDkk-1-derived Synthetic Peptides and Lithium Chloride for the Control and

doi: 10.1074/jbc.M406275200 originally published online October 25, 20042005, 280:2309-2323.J. Biol. Chem. 

  10.1074/jbc.M406275200Access the most updated version of this article at doi:

 Alerts:

  When a correction for this article is posted• 

When this article is cited• 

to choose from all of JBC's e-mail alertsClick here

  http://www.jbc.org/content/280/3/2309.full.html#ref-list-1

This article cites 33 references, 10 of which can be accessed free at

by guest on June 7, 2020http://w

ww

.jbc.org/D

ownloaded from