In Vivo and In Vitro Characterization of Insulin-Producing ......Cytospin slides from differentiated mBMDS (D-mBMDS) cells were made for insulin and C-peptide protein expression. The

In Vivo and In Vitro Characterization ofInsulin-Producing Cells Obtained From MurineBone MarrowDong-Qi Tang, Li-Zhen Cao, Brant R. Burkhardt, Chang-Qi Xia, Sally A. Litherland,

Mark A. Atkinson, and Li-Jun Yang

Efforts toward routine islet cell transplantation as ameans for reversing type 1 diabetes have been ham-pered by islet availability as well as allograft rejection.In vitro transdifferentiation of mouse bone marrow(BM)-derived stem (mBMDS) cells into insulin-produc-ing cells could provide an abundant source of autolo-gous cells for this procedure. For this study, we isolatedand characterized single cell-derived stem cell linesobtained from mouse BM. In vitro differentiation ofthese mBMDS cells resulted in populations meeting anumber of criteria set forth to define functional insulin-producing cells. Specifically, the mBMDS cells ex-pressed multiple genes related to pancreatic �-celldevelopment and function (insulin I and II, Glut2, glu-cose kinase, islet amyloid polypeptide, nestin, pancre-atic duodenal homeobox-1 [PDX-1], and Pax6). Insulinand C-peptide production was identified by immunocy-tochemistry and confirmed by electron microscopy. Invitro studies involving glucose stimulation identifiedglucose-stimulated insulin release. Finally, these mB-MDS cells transplanted into streptozotocin-induced di-abetic mice imparted reversal of hyperglycemia andimproved metabolic profiles in response to intraperito-neal glucose tolerance testing. These results indicatethat mouse BM harbors cells capable of in vitro trans-differentiating into functional insulin-producing cellsand support efforts to derive such cells in humans as ameans to alleviate limitations surrounding islet celltransplantation. Diabetes 53:1721–1732, 2004

Type 1 diabetes is an insulin-dependent, autoim-mune disorder characterized by the destructionof insulin-producing �-cells (1). Hence, a rever-sal of type 1 diabetes could be afforded by

replacement of functional �-cells. Unfortunately, islettransplantation has historically been hampered by immune

rejection and/or a recurrent attack by underlying autoim-munity against islets, as well as the scarcity of donor islets(2,3). One theoretical alternative for islet transplantationwould involve the use of a renewable source of stem cellscapable of self-renewal and differentiation, as well as thatof insulin production. Indeed, the development of a sim-ple, reliable procedure to obtain autologous stem cellshaving the ability to differentiate into functional insulin-producing cells would provide a potentially unlimitedsource of islet cells for transplantation and alleviate themajor limitations of availability and allogeneic rejection.

Recent studies have shown that bone marrow (BM)-derived stem (BMDS) cells have the ability to differentiateinto a number of neuroectodermal, endothelial, mesenchy-mal, epithelial, and endodermal cell types (4–10). Theability of hepatic stem cells, such as oval cells andfunctional hepatocytes, to derive from BM cells has alsobeen suggested in several in vivo (11–13) and in vitro (10)studies. The observation that human BMDS cells candifferentiate into mature hepatocytes (14,15) confirms theclose interrelationship of BMDS cells and hepatocytes.Previously, we demonstrated (16) that highly purified rathepatic oval cells can be induced to differentiate intofunctional insulin-producing cells when cultured long termin a high-glucose environment, that these differentiatedoval cells express insulin, glucagon, and pancreaticpolypeptide, and that they respond (i.e., produce insulin)to a high-glucose challenge. A key question that remainedfollowing those studies was whether BMDS cells could beinduced to become functional insulin-producing cells. Be-cause the pancreas and liver have common precursor cellsduring embryogenesis (17), stem cells of these two organsmay have the same origin, that being from BM.

In this present study, we isolated murine BMDS (mBMDS)cells and obtained single-cell–derived cell clones thatwere subsequently induced to transdifferentiate into insu-lin-producing cells under culture conditions containinghigh concentrations of glucose and the addition of �-cell–stimulating factors. The functionality of these cells wasconfirmed by insulin production and release in a glucose-responsive manner and by their reversal of hyperglycemiaafter being transplanted into mice rendered diabetic bytreatment with streptozotocin (STZ). Taken together, ourresults indicate that under suitable conditions, mBMDScells can be induced in vitro to differentiate into functionalinsulin-producing cells capable of normalizing hyperglyce-mia in a diabetic animal model. This study provides

From the Department of Pathology, Immunology, and Laboratory Medicine,University of Florida College of Medicine, Gainesville, Florida.

Address correspondence and reprint requests to Dr. Li-Jun Yang, 1600 SWArcher Rd., P.O. Box 100275, Gainesville, FL 32610. E-mail: [email protected].

Received for publication 2 October 2003 and accepted in revised form 25March 2004.

D.-Q.T. and L.-Z.C. contributed equally to this work.BM, bone marrow; DAPI, 4�,6-diamidino-2-phenylindole; D-mBMDS, differ-

entiated murine BM-derived stem cells; ELISA, enzyme-linked immunosorbentassay; GLP, glucagon-like peptide; IPGT, intraperitoneal glucose tolerance;mBMDS, murine BM-derived stem cells; Pdx, pancreatic duodenal homeobox;STZ, streptozotocin.

© 2004 by the American Diabetes Association.

DIABETES, VOL. 53, JULY 2004 1721

support for continuing efforts aimed at utilizing adult stemcells as a steady and renewable source of autologousinsulin-producing cells for transplantation in patients withtype 1 diabetes.

RESEARCH DESIGN AND METHODS

BM isolation and derivation of single-cell–derived stem cell clones. Balb/cmice were purchased from the mouse production facility in the Department ofPathology, University of Florida. All procedures were performed underprotocols approved by the Institutional Animal Care and Use Committee at theUniversity of Florida. BM was obtained from the femurs and tibias (longbones) of 10 male Balb/c mice. The bones were sterilized by immersion in 70%ethanol, followed by removal of the remaining skin and muscles. BM wasexposed by cutting the ends of the bones and extruded by inserting a needleand forcing cell culture medium with 10% FCS (HyClone, Logan, UT) throughthe bone shaft. Gentle pipetting resulted in the generation of a single-cellsuspension. In such efforts, one mouse would routinely yield �5 � 106 totalBM cells. BM cells (2 � 106/ml) were cultured (37°C, 5% CO2) in 12-well plateswith RPMI 1640 medium containing 10% FCS, 20 mmol/l HEPES, 1� penicillin,and streptomycin (Life Technology, Grand Island, PA). One week later, totalmedia in the culture was changed and included removal of floating cells. Threeweeks later, adherent cells gaining 80% confluence were passaged by pipettingwith or without the use of trypsin. Following three to four passages, the cellsbecome morphologically homogeneous, with a slim-spindle appearance. Atthis stage, single-cell–derived BMDS cell clones were derived using a cloningcylinder (Fisher Scientific, Pittsburgh, PA), with the selected cells beingexpanded and used for characterization of stem cell properties and for studiesinvolving in vitro differentiation.Antibodies. Rabbit anti-insulin polyclonal IgG (Santa Cruz Biotechnology,Santa Cruz, CA), guinea pig anti-insulin (Dako, Carpinteria, CA), rabbit anti-ratC-peptide antibody (Linco Research, St. Charles, MO), anti-rabbit IgG, guineapig serum, and Cy3-coupled anti–guinea pig IgG (RDI, Research Diagnostics,Flanders, NJ) were obtained and utilized as indicated and in accordance withthe manufacturer’s recommendations. Antibodies directed against CD34,CD45, C-kit, and Sca-1 (BD Pharmingen, San Diego, CA) were used for theflow cytometric analysis.Flow cytometric analysis. The mBMDS cells at three to four passages werereleased by trypsinization. The cells were incubated with anti-mouse fluores-cent dye–labeled hematopoietic antibodies, with 10,000 events acquired foranalysis of fluorescence intensity, as previously described (18). Isotype-matched mouse immunoglobulins served as controls for autofluorescence.In vitro differentiation cultures. To induce the mBMDS cells to undergopancreatic endocrine cell differentiation, the cloned cells were cultured (37°C,5% CO2) in basic medium composed of RPMI 1640 medium (10% FCS) for 2–4months in the presence of low (5.5 mmol/l) or high (23 mmol/l) concentrationsof glucose. Cellular differentiation was monitored by observation of three-dimensional, islet-like cell cluster formation and by the expression of genesrelated to pancreatic �-cell development and insulin production. To promotecellular maturation, the cells were cultured (37°C, 5% CO2) for 7 days in RPMI1640 medium containing 5.5 mmol/l glucose, 5% FCS, and 10 mmol/l nicotin-amide (Sigma, St. Louis, MO). The cells were then cultured for an additional5–7 days in the presence of 10 nmol/l exendin 4 (Sigma).Cell line culture. The rat INS-1 cell line (clone 832/13), a cell line capable ofinsulin release in response of glucose stimulation, was a generous gift fromDr. Christopher B. Newgard (Duke University, Durham, NC). This cell line wasderived from stable transfection of a plasmid containing the human proinsulingene driven by a cytomegalovirus promoter and has the capacity to expressand process both rat and human insulin. The cells were maintained in RPMI1640 medium with 11.1 mmol/l D-glucose supplemented with 10% fetal bovineserum, 100 units/ml penicillin, 100 �g/ml streptomycin, 10 mmol/l HEPES, 2mmol/l L-glutamine, 1 mmol/l sodium pyruvate, and 50 �mol/l �-mercaptoetha-nol at 37°C/5% CO2 in a humidified atmosphere (19). This cell line was used asa positive control for studies of insulin content and insulin release.RT-PCR. Total RNA was prepared from BMDS cell cultures maintained inlow- or high-glucose culture for 4 months using TRIzol reagent. To eliminategenomic DNA contamination, mRNAs were purified using oligo-dT cellulose(Micro-FastTrack 2.0 Kit; Invitrogen, Carlsbad, CA) according to the manu-facturer’s protocol. Transcriptional gene expression related to pancreaticendocrine development and function as well as other lineage markers(neuronal, intestine, and liver) from these cultures was determined by RT-PCRaccording to a published protocol (16) with minor modifications. The forwardand reverse primers of each PCR set were designed to be located in differentexons based on sequences obtained from GenBank to distinguish the PCRproducts from DNA contamination. Key PCR products of genes related topancreatic development were confirmed by sequence analysis. The name and

sequences of the primers, the sizes of PCR products, cycles, and annealingtemperature for each pair are listed in Table 1.DNA-PCR for detection of microsatellite polymorphism. Genomic DNAwas isolated from tissues and cell lines with phenol-chloroform followed byethanol precipitation. Extracted DNA was resuspended in TE buffer (10mmol/l Tris-Cl, pH 7.5, 1 mmol/l EDTA). Five highly polymorphic markersunique for mice, including D2Mit30, D3Mit15, D6Mit15, D11Mit4, and D2Nds3,were selected from the Mouse Genome Informatics database. The sequencesfor the microsatellite markers can be obtained from The Jackson Laboratorieswebsite (http://www.informatics.jax.org). DNA (500 ng) was used as a tem-plate in PCRs. The sizes of the DNA products obtained with the five markers(D2Mit30, D3Mit15, D6Mit15, D11Mit4, and D2Nds3) in Balb/c mice are 136,143, 195, 242, and 400 bp, respectively.Immunocytochemistry and immunofluorescence. Cytospin slides fromdifferentiated mBMDS (D-mBMDS) cells were made for insulin and C-peptideprotein expression. The cells were fixed with 4% formaldehyde for 30 min atroom temperature, and immunocytochemistry performed with polyclonalguinea pig anti-insulin (1:500) (Dako) and guinea pig anti-rat C-peptideantibody (1:100) (Linco Research) for 1 h. After washing, the cells wereincubated with Cy3-coupled anti-guinea pig (1:1,000) secondary antibodies(RDI) for 30 min. Guinea pig serum was used as a negative control. Cells werethen examined by fluorescence microscopy (Olympus BX51).Deconvolution microscopy. Cells were stained with Cy3-conjugated second-ary antibodies after they were incubated with antibodies specific for insulin orC-peptide. The nuclei were counterstained with 4�,6-diamidino-2-phenylindole(DAPI) and imaged by deconvolution microscopy using an Olympus OMTinverted fluorescent microscope system equipped with Delta Vision deconvo-lution analysis software. The images depict three-dimensional projections of25 optical slices (each 0.2-�m thick) through the cell, with the center focusedon the DAPI-stained chromatin in the nuclei. All images used in this reportwere scale adjusted, including images of staining with nonspecific isotypeantibody conjugates as a negative control.Mouse insulin enzyme-linked immunosorbent assay. D-mBMDS cellswere cultured (37°C, 5% CO2) in the presence or absence of exendin-4 for 7days after 1 week of 10 mmol/l nicotinamide treatment in RPMI 1640containing 5% fetal bovine serum and 5.5-mmol/l glucose. Following this, thecells were confirmed to express insulin genes by RT-PCR. In a parallelexperiment, the cells were cultured in the presence of exendin 9-39 for 7 days.The cells were switched to serum-free medium containing 0.5% BSA for 12 h,washed twice with PBS, then stimulated by the addition of 23 mmol/l glucosefor 2 h. The culture media was collected and frozen at �70°C until assay forinsulin release. Importantly, serum-free culture medium containing 0.5% BSAwas used as a control for secreted insulin measurements. Insulin release wasdetected by using an ultrasensitive mouse insulin enzyme-linked immunosor-bent assay (ELISA) kit (Alpco Diagnostics, Windham, NH) following themanufacturer’s protocols. According to the manufacturer’s instructions, thisassay does not detect proinsulin.Electron microscopy with immunogold labeling. For immunogold local-ization of insulin, the cells were embedded in Lowicryl K4M resin (EMSciences, Fort Washington, PA). Ultrathin sections were blocked with 5%BSA/5% normal goat serum in PBS and then incubated overnight at 4°C inrabbit anti-insulin antibody (Santa Cruz Biotechnology) diluted 1:50 in PBScontaining 0.2% BSA and 10 mmol/l NaN3. After washing, the samples wereincubated for 1.5 h at room temperature with the secondary goat anti-rabbitIgG antibody conjugated to 0.8-nm colloidal gold particles (Aurion EM GradeUltra Small, EM Sciences), washed, treated with 1.25% glutaraldehyde in PBS,and washed again. The gold particles were silver enhanced for 45 min at roomtemperature (Aurion R-Gent SE; EM Sciences). The samples were counter-stained using uranyl acetate and lead citrate, then viewed using a ZeissEM-10A transmission electron microscope.Transplantation studies in mice. Balb/c male mice received two intraperi-toneal injections of STZ at 250 and 50 �g/g body wt, 3 days apart, accordingto published procedures (20,21) with minor modification. Blood glucose levelswere monitored using an AccuChek glucose detector (Roche Diagnostics,Indianapolis, IN). Within 12 days of the first injection, all Balb/c mice becamehyperglycemic, with blood glucose levels �350 mg/dl. The D-mBMDS cells(5 � 106/mouse) were transplanted into the left renal capsule and the distal tipof the spleen of six diabetic mice. Five diabetic mice received sham surgerywithout implants as a control. The nonfasting blood glucose levels weremonitored at 1600 every 2 days following transplantation. Most of the diabeticmice with sham surgery died between 15 and 20 days because they did notreceive insulin treatment. The diabetic mice with D-mBMDS cell transplantswere killed 26 days after transplantation. The pancreas tissue was harvestedfor morphologic analysis. For the intraperitoneal glucose tolerance (IPGT)test, normal nondiabetic Balb/c male mice (n � 5) and diabetic mice (n � 3)with normalized glucose levels following the D-mBMDS cell transplantation

FROM BONE MARROW TO INSULIN-PRODUCING CELLS

1722 DIABETES, VOL. 53, JULY 2004

TA

BLE

1List

ofm

ousegene–specific

primers

inR

T-P

CR

analysisof

cells

Genes

Forw

ardprim

erR

everseprim

erSize

ofP

CR

product(bp)

GenB

ankaccession

no.A

nnealingtem

perature(oC

)N

o.of

cycles

InsulinI

TG

GG

GG

TC

GG

GA

AT

CA

CT

GG

TT

GG

GC

CT

TA

GT

TG

CA

GT

AG

TT

396X

0472560

32Insulin

IIC

TG

GC

CC

TG

CT

CT

TC

CT

CT

GG

CT

GA

AG

GT

CA

CC

TG

CT

CC

CG

G204

NM

000838758

32G

lut-2C

AT

TC

TT

TG

GT

GG

GT

GG

CC

CT

GA

GT

GT

GT

TT

GG

AG

CG

221X

1698655

35G

lucokinaseG

CA

GA

TC

CT

GG

CA

GA

GT

TC

CA

GG

AA

GG

AG

AA

GG

TG

AA

GC

CC

A408

BC

001113966

32IA

PP

CC

TC

AT

CC

TC

TC

TG

TG

GC

AC

CA

CG

TT

GG

TT

GG

TG

GG

AG

175M

2538955

32N

estinG

GA

GA

GT

CG

CT

TA

GA

GG

TG

CG

TC

AG

GA

AA

GC

CA

AG

AG

AA

G327

NM

01670158

35P

dx-1T

GG

AT

AA

GG

GA

AT

TG

CT

TA

AC

CT

TT

GG

AA

CG

CT

CA

AG

TT

TG

TA

249N

M008814

6232

Pax6

GC

AC

AC

GC

CC

TG

GT

TG

GT

CA

CT

GT

AC

GT

GT

TG

GT

GA

G512

NM

01362760

32P

ax4G

GA

CT

CT

TT

GT

GA

AT

GG

CC

GG

TT

TA

GC

TG

GG

CA

AT

TC

GA

GC

C236

XM

13302364

35N

euroDA

TG

AC

CA

AG

GC

GC

GC

CT

AG

AA

CA

GG

AC

AG

TC

AC

TG

TA

CG

CA

C425

U28068

5535

Oct-4

GG

CG

TT

CT

CT

TT

GG

AA

AG

GT

GT

TC

CT

CG

AA

CC

AC

AT

CC

TT

CT

CT

312X

M285447

5635

Isl-1T

TT

CC

CT

GT

GT

GT

TG

GT

TG

GT

CT

TC

TC

GG

GC

TG

TT

TG

T501

NM

02145956

35G

LP-1R

GA

AT

AC

CG

GC

GG

CA

GT

GC

CA

CT

GT

GC

AA

GT

GT

CT

GA

AG

CC

A402

NM

_02133256

35A

lbumin

AT

GC

TC

AT

AC

GA

TG

AG

CA

TG

CA

TG

GT

GG

CA

GG

CT

GG

GG

TT

G245

BC

02464356

35T

TR

TC

GC

TG

GA

CT

GG

TA

TT

TG

TG

GT

TG

GC

TG

TG

AA

AA

CC

AC

AT

CC

323B

C024702

5635

AF

PC

GT

GA

CG

GA

GA

AG

AA

TG

TG

CT

CT

TA

AT

TC

CT

TT

GC

AA

TG

GA

515B

C066206

5635

GF

AP

CT

AA

GA

TG

AA

GT

TA

TG

GG

AT

GA

CA

TT

TA

AG

TG

TA

TG

GC

AG

T389

NM

_01027556

35T

ubulinT

CT

GG

GA

GG

TC

AT

CA

GC

GA

TT

CA

CG

CA

CC

TT

GC

TG

TG

AG

CA

412N

M_023279

5635

NF

MT

AT

GC

TC

AG

CT

CG

GC

CG

AG

AG

GC

AC

TT

GA

GC

CT

TC

TC

GT

GG

T311

NM

_00869156

35M

uc2G

GC

AT

TG

TG

TG

CC

AA

CC

AA

AG

CC

TT

GG

GC

AC

AC

AG

GA

AT

AA

AC

TG

235X

M_133960

5635

SucraseA

CG

AT

AA

TA

GC

TA

TC

GC

TC

TT

AA

AG

AT

TG

GC

CA

TG

TT

TT

CC

693X

M_143332

5635

Villin

GG

CT

AT

GC

AG

AT

GG

TA

CC

TG

TA

GT

CG

CT

GG

AC

AT

CA

CA

GG

A341

NM

_00950956

35

AF

P,

fetal

protein;G

FA

P,

glialfibrillary

acidicprotein;

IAP

P,

isletam

yloidpolypeptide;

MU

C2,

mucin

2.

D.-Q. TANG AND ASSOCIATES


received intraperitoneal injections of glucose (2 mg/g body wt) according topublished procedures (20). Blood glucose levels were monitored at 0, 30, 60,90, 120, and 150 min for each mouse.Measurement of apoptosis. The mBMDS cells were cultured in mediumcontaining 23 mmol/l glucose for various times including 1 week, 2 weeks, 1month, and 2 months. Cells were passaged when they reached 80–90%confluency. Cultured mBMDS cells in the medium containing a 5.5-mmol/lglucose concentration served as a baseline control for no treatment. Cellswere released from culture dishes by trypsinization and incubated in the samemedium for an additional 2 h in suspension at 37°C. The cells them were thensubjected to testing for apoptosis using an Annexin V-PE apoptosis detectionkit I (BD Biosciences Pharmingen) following the manufacturer’s protocol.Statistics. Evidence of statistical significance was determined by Fisher’sexact testing. A P value of 0.05 was deemed significant.

RESULTS

Derivation and characterization of BMDS cells from

mice. BM cells from 10 Balb/c male mice were used toobtain BMDS cells. Adherent BMDS cells were derivedfrom cultures of unsorted BM cells. The unattached cellswere removed following 2–7 days of culture, with spindle-shaped adherent cells (Fig. 1A, upper panel) cultured foran additional 2–3 weeks, until the spindle-shaped adherentcells reached 70–80% confluence. The cells were thenreleased from the substrate surface with trypsin-EDTA orpipetting and were replated under the same culture con-ditions for several passages. A single-cell–derived cellclone was obtained by trypsinization of a single-cell–derived cell cluster using a cloning cylinder. The single-

cell–derived mBMDS cells, as well as the mixed mBMDScells, were then characterized for their stem cell proper-ties and their phenotypes. Specifically, the resulting celllines were assessed by flow cytometric analysis of surfacemarkers and by differentiation potential under selectiveculture conditions. The phenotype of these mBMDS cellswas predominantly negative for CD45, CD34, C-kit, andSca-1, whereas a small subpopulation of these cells waspositive for Sca-1 (Fig. 1B). Thus, the immunophenotypeappeared similar but not identical to that of multipotentadult progenitor (MAP) cells reported by Jiang et al. (22).We observed that these mBMDS cells could be induced todifferentiate into neural or endothelial cells (data notshown), but induction to hepatocytes using the protocolsof Schwartz et al. (10) was not successful. These findingsare consistent with the notion that mBMDS cells possessstem cell properties. Therefore, we believe that the BMDScells most likely represent pluripotent mesenchymal stem/precursor cells.In vitro differentiation of mBMDS cells into insulin-

producing cells. To induce cell differentiation, six single-cell–derived clones of the mBMDS cells were switchedinto RPMI 1640 medium containing 10% FCS and a high-glucose concentration (23 mmol/l). After 2–4 months of invitro induction, four of the six single-cell–derived clonalcultures began to form three-dimensional clusters similar

FIG. 1. Isolation, derivation, and characterization of clonal mBMDS cells. A: BM cells (2 � 106 cells/ml) from Balb/c mice were plated and culturedfor 2–7 days to obtain the adherent mBMDS cells (top). Cloned mBMDS cells were used for the in vitro differentiation protocol by culturing cellsin the presence of a 23-mmol/l glucose concentration for various times. Many cell clusters at various stages of cluster formation during the courseof induction of cell differentiation were observed, with a representative shown in the bottom panel. HG, high glucose; LG, low glucose. B: Arepresentative phenotype of the mBMDS cells. FITC, fluorescein isothiocyanate; PE, phycoerythrin.



to that shown in our previous study (16) (Fig. 1, lower

panel). To promote further differentiation, the cells wereswitched to medium containing 5% FCS with 10 mmol/lnicotinamide, 10-nmol/l exendin 4, and 5.5 mmol/l glucose.These culture conditions promoted further cluster forma-tion, in which the cell clusters increased in both numberand mass, as well as increasing their glucose responsive-ness (data not shown). Cellular differentiation was thenmonitored by RT-PCR for ascertainment of gene expres-sion (see below).Gene expression of mBMDS and D-mBMDS cells. Todetermine whether the mBMDS cells had undergone pan-creatic differentiation, gene expression profiles for pancre-atic �-cell differentiation markers and hormones wereassessed using RT-PCR. Four of the six mBMDS clonesunderwent pancreatic endocrine differentiation as evi-denced by expression of Pdx-1 and insulin genes. Asillustrated in Fig. 2, cells cultured under high-glucoseconcentrations (23 mmol/l) for 4 months expressed multi-ple genes characteristic of endocrine �-cell development,including insulin I and II, Glut-2, glucose kinase, isletamyloid polypeptide, nestin, Pdx-1, and Pax6. However,gene expression of Pax4, NeuroD, and islet-1 (Fig. 2, upper

panel) was not detected. The Oct-4 gene, typical forpluripotent embryonic stem cells, was not detected inmBMDS or D-mBMDS cells. In contrast, mBMDS cellscultured under a low-glucose concentration (5.5 mmol/l)for 4 months expressed no detectable levels of the afore-mentioned genes with the exception of nestin (Fig. 2,lower panel). Further differentiated mBMDS cells (latestage), similar to the mouse �-cell line (�-TC) expressedthe glucagon-like peptide (GLP)-1 receptor gene (Fig. 3D),yet this gene expression was not observed in undifferen-tiated mBMDS cells or in early stages of the D-mBMDScells (data not shown).

To determine whether any non–�-cell markers wereexpressed in these cells, several markers representinghepatic, intestinal, and neuronal differentiation were ana-lyzed by RT-PCR to compare gene expression between theundifferentiated and D-mBMDS cells. Mouse liver, brain,and intestinal tissue served as a positive control. Asindicated in Fig. 3, no detectable expression of the hepatic(Fig. 3A), intestinal (Fig. 3B), and neuronal (Fig. 3C) geneswas observed in either differentiated or undifferentiated

cells, with the exception of nestin. These results suggestthat culture of mBMDS cells under a long-term, high-glucose condition favors their pancreatic endocrine differ-entiation.

To confirm that the D-mBMDS cells are indeed derivedfrom Balb/c mice and that there is no cross-contaminationfrom different cell lines or species, we analyzed cellularDNA for five mouse microsatellite molecular markers:D2Mit30, D3Mit15, D6Mit15, D11Mit4, and D2Nds3. Tis-sues from Balb/c mice as well as rat and human wereobtained for these analyses. Genomic DNA from Balb/cmice served as a positive control in these analyses. Thesemouse microsatellite markers have the ability to distin-

FIG. 2. Gene expression patterns in D-mBMDS cells (high glucose, HG) and undifferentiated mBMDS cells (low glucose, LG). Total RNA wasisolated from cultures of undifferentiated mBMDS (grown in LG medium for 4 months) and D-mBMDS (grown in HG medium for 4 months) cellcultures. RT-PCR was performed to detect genes related to �-cell development and insulin production. All PCR products were verified by DNAsequence analysis. The results were repeated at least three times.

FIG. 3. Gene expression of nonpancreatic genes. Total RNA wasisolated from cultures of undifferentiated (Un-D) and D-mBMDS (Dif)cell cultures. RT-PCR was performed to detect genes related to hepatic(A), neuronal (B), and intestinal (C) differentiation. GLP-1 receptorgene expression among cells (D) is presented. AFP, � fetal protein;GFAP, glial fibrillary acidic protein; MM, molecular marker; MUC2,mucin 2; NFM, neurofilament protein.



guish tissues among different species (mouse, rat, andhuman), but in addition, can also be used to distinguishdifferent strains within the same mouse species (i.e.,NOD/MrKTac, Balb/cJ, C57BL/6J CAST, etc.) due to poly-morphisms at a particular site of the host chromosome.The PCR results assigning the specific polymorphism (Fig.4A) showed that the sizes of the PCR products of the fivemarkers (D2Mit30, D3Mit15, D6Mit15, D11Mit4, andD2Nds3) were precisely located at 136, 143, 195, 242, and400 bp, respectively, and that the pattern was identical tothat of a Balb/c mouse (Fig. 4B, left panels). There was nocross-contamination among species present (Fig. 4B).Insulin and C-peptide synthesis by D-mBMDS cells. Todetermine whether the D-mBMDS cells actually synthesizeinsulin protein and release C-peptide, the differentiated

cells after further culture (1 week) under conditionsincluding the presence of nicotinamide and exendin 4,were stained with anti-insulin and anti–C-peptide antibod-ies and visualized by fluorescence microscopy (Fig. 5A).Figure 5A shows that �10–20% of the D-mBMDS cells(counted in various microscopic fields) stained stronglywith the anti-insulin antibody (bottom middle panel).Strong C-peptide cytoplasmic staining was also detected in10–20% of the examined cells (Fig. 5A, bottom right

panel). The staining pattern and intensity for both insulinand C-peptide was similar to that observed in the positivecontrol INS-1 cells (Fig. 5A, top panels). Matched isotypecontrol antibodies served as negative controls (Fig. 5A,left). These results indicated that the D-mBMDS cells areindeed synthesizing and processing insulin as indicated by

FIG. 4. Confirmation of the mBMDS cell origin by microsatellite. Five mouse-specific probes were selected for studies of polymorphism todistinguish among species and among mouse strains. The sizes of the PCR products of the five markers (D2Mit30, D3Mit15, D6Mit15, D11Mit4,and D2Nds3) in Balb/c mice are 136, 143, 195, 242, and 400 bp, respectively.

FIG. 5. A: Insulin and C-peptide production by D-mBMDS cells. Cytospin slides made of cultured D-mBMDS and INS-1 cells were stained withanti-insulin (middle) and anti–C-peptide (right) antibodies and visualized under fluorescence microscopy. INS-1 cells were used as a positivecontrol for insulin and C-peptide. A negative control utilizing isotype-matched antibodies is shown (left). B: Analysis of insulin granules bydeconvolution microscopy. The distribution of insulin and C-peptide granules in both INS-1 and D-mBMDS cells visualized by deconvolutionmicroscopy following insulin and C-peptide immunostaining. Insulin and C-peptide molecules were stained in red color, and the nuclei werestained with DAPI (blue color). C: Analysis of insulin granules by electron microscopy. Immunogold labeling of insulin in INS-1 (left panel) andin D-mBMDS (right panel) cells is shown. Arrows indicate immunogold-labeled insulin granules.



the presence of C-peptide, a byproduct of de novo insulinrelease and one observed following further in vitro differ-entiation into more mature insulin-producing cells.Analysis of insulin granules by deconvolution and

electron microscopy with gold labeling. To confirmand identify the distribution of insulin granules in singlecells, we used immunohistochemical staining with anti-insulin antibodies visualized by deconvolution and elec-tron microscopy to compare the D-mBMDS cells withINS-1 cells in terms of their granular distribution andultrastructure. Figure 5B demonstrates a similar distribu-tion of insulin granules in the D-mBMDS cells to thatobserved in the INS-1 cells. Figure 5C confirms that theglobular structures observed in both INS-1 (Fig. 5C, left)and D-mBMDS (Fig. 5C, right) cells contain insulin (highelectron-dense core in the granules are indicated by ar-row), but in the case of mBMDS cells, at levels far lessthan that observed in INS-1 control cells.Insulin content and release in response to glucose

stimulation. To determine whether the D-mBMDS cellswere responsive to a glucose challenge, insulin releasefrom undifferentiated and D-mBMDS cells was measuredusing an ultrasensitive mouse insulin ELISA. In order toenhance the sensitivity of these cells to high-glucose

challenge, the differentiated cells were switched to low-serum, low-glucose medium plus nicotinamide for 1 week.Since the GLP-1R gene was expressed in the latter stagesof D-mBMDS cells, as shown in Fig. 3D, we then culturedcells in the presence of either exendin 4, or its antagonist,exendin 9-39, for 7 days before analysis. The cells werethen washed twice with PBS and switched to serum-freelow-glucose medium containing 0.5% BSA (overnight incu-bation), then stimulated by the addition of 23 mmol/lglucose for 2 h. The cell culture media was then collectedfor the analysis. All studies were performed in triplicate.The results shown in Fig. 6 indicate that approximatelyfourfold more insulin was released in exendin 4–treatedcells in comparison to the cells treated with exendin 9-39under long-term high-glucose conditions. In contrast, con-trol mBMDS cells cultured in low concentrations of glu-cose (5.5 mmol/l) showed no significant release of insulinin the presence or absence of glucose challenge, even inthe presence of exendin 4. These data suggest that high-glucose culture plays an indispensable role in the transdif-ferentiation of BMDS cells into insulin-producing cells andthat differentiated BMDS cells were responsive to glucosechallenge. Moreover, the results also indicated that theD-mBMDS cells might represent a precursor to �-like cells

FIG. 5—Continued.



and that further induction might be needed to reach a highdegree of differentiation and maturation, as would beobserved in the in vivo hyperglycemic environment ofdiabetic animals.

As an additional assessment, we quantitatively evalu-ated the content of insulin in both mBMDS-derived insulin-producing cells as well as in the INS-1 cell line (clone832/13) (Table 2). The ratio of insulin content to insulinrelease did not vary dramatically between the D-mBMDScells (8.7) and INS-1 cells (11.5), although the amount ofinsulin content and release in INS-1 cells was significantlyhigher. Because the rat insulinoma INS-1 cell line (clone832/13) was derived from stable transfection of a plasmidcontaining the human proinsulin gene driven by a cyto-megalovirus promoter, it must be remembered that thecontent and release of insulin in INS-1 cells does notrepresent the physiological levels seen in native pancreatic�-cells.Reversal of hyperglycemia in STZ-induced diabetic

mice. To determine whether the D-mBMDS cells pos-sessed the capacity to correct hyperglycemia in diabeticmice, Balb/c mice were STZ induced to become diabeticbefore transplantation with D-mBMDS cells (5 � 106/mouse). Five STZ-induced diabetic mice received shamsurgery without cellular implantation as a control. Asdemonstrated in Fig. 7A, glucose levels in the mBMDScell–implanted mice decreased and normalized within 1week following transplantation. In contrast, blood glucoselevels in the diabetic control mice remained elevated (P 0.01; diabetes frequency at day 14). The diabetic control

mice lost weight persistently (data not shown), with manydying between 15 and 20 days after diabetes onset. Theseresults suggested that the D-mBMDS cells are functional invivo and capable of reversing hyperglycemia in diabeticmice. To further evaluate the function of the implantedmBMDS cells, we performed an IPGT test on control(nondiabetic) Balb/c mice (n � 5) and transplanted Balb/cmice (n � 3) after 14 days of normalized glucose levelsfollowing the transplant. These studies involved glucoseadministration (2 mg/g) provided under fasting conditions.As illustrated in Fig. 7B, blood glucose levels in normalcontrol mice rose rapidly, with peak values obtained at 30min, followed by a return to the normal range between 60and 90 min. Blood glucose levels in the implanted micewere generally higher, but likewise displayed a peak at 30min and returned to the normal range at 150 min. Thisresult indicated that the implanted mBMDS cells wereindeed responsive to a glucose challenge in vivo, but werenot as effective as native pancreatic �-cells in terms ofrestoring normoglycemia.Effects of high-glucose culture on mBMDS cells.

These results demonstrated that long-term culturing ofadult stem cells under high-glucose conditions promotedtheir pancreatic endocrine differentiation. However, thepotential apoptotic effect of high-glucose cultures on themBMDS cell was unknown. To evaluate this possibility,the mBMDS cells were cultured under normal-glucose (5.0mmol/l) or high-glucose (23 mmol/l) conditions for varioustime periods for up to 2 months, and then the cells weresubjected to apoptosis analysis as described in RESEARCH

DESIGN AND METHODS. As shown in Fig. 8, the high-glucosecultures indeed induced apoptosis in mBMDS cells, eventhough overt apoptotic changes were not visible by lightmicroscopy. However, the percentages of the cells under-going apoptosis remained relatively constant, between 34and 37% in all culture conditions. Indeed, there were nostatistical differences between short-term and long-termhigh-glucose cultures in that parameter. However, therewas a marked difference in the increased number ofapoptotic cells present in the high-glucose cultures whencompared with their normal culture controls. Interest-ingly, at least 60% of cells in all high-glucose cultureconditions did not show signs of either apoptosis ornecrosis.

DISCUSSION

Recent studies have demonstrated the feasibility of gener-ating insulin-producing cells obtained from progenitorcells of various cellular sources, including the pancreas(23,24), liver (16), and intestinal epithelium (25), as well asthe pluripotent embryonic stem cells of mouse (20,26) andhuman (27) origin. However, even with the conceptualadvances offered by these findings, some obstacles, suchas immune rejection and autoimmunity against newlyformed �-cells derived from pancreatic stem cells, stillremain. Despite their promising potential, it may alsoprove difficult to obtain enough autologous adult stemcells from these organs.

To overcome these limitations, we explored the possi-bility of using human and mouse BMDS cells as sourcesfor transdifferentiation into insulin-producing cells underspecific in vitro culture conditions. BM has been known

FIG. 6. Insulin release of mBMDS cells (�), D-mBMDS cells treatedwith exendin 4 (f), and D-mBMDS cells treated with exendin 9-39 (u)upon glucose stimulation. Cells were cultured in RPMI 1640 mediumcontaining 5.5 mmol/l glucose and 5% FCS, plus nicotinamide for 1week and for an additional 1 week in the presence of either exendin 4or exendin 9-39. The cells were then switched to serum-free mediumcontaining 0.5% BSA for 12 h and then stimulated with 23 mmol/lglucose for 2 h. The cell culture medium was then collected for assay ofinsulin release. Released insulin in the media was detected by anultrasensitive ELISA kit. Results shown here represent those of threeseparate experiments.

TABLE 2Comparison of insulin release and content between mBMDS andINS-1 cells

Cell type

Insulinrelease(ng/ml)

Insulin content(ng/ml)

Content/release

ratio

INS-1 98.3 � 18.31 1,130.3 � 204.23 11.5D-mBMDS 1.5 � 0.17 13.3 � 1.69 8.7

Data are means � SD.



for years to represent a safe and abundant source for largequantities of adult stem cells. In the present study, weisolated, cloned, and characterized mBMDS cells. We alsogenerated functional insulin-producing cells from the mB-MDS cells under an in vitro differentiation procedure andconfirmed the presence of insulin production by RT-PCR,immunofluorescence, and electron microscopy combinedwith gold anti-insulin labeling. Furthermore, we tested thefunctionality of the in vitro–generated insulin-producing cellsfrom mouse BM by measuring insulin release in response toa glucose challenge and by demonstrating a reversal ofdiabetes upon subsequent implantation of these cells intodiabetic mice. In addition to these studies, we have derivedislet-like, functional insulin-producing cells from humanBMDS cells (D.-Q.T., B.R.B., L.-Z.C., S.A.L., M.A.A., L.-J.Y.,unpublished data). Taken collectively, these studies providedirect evidence that the BM contains pluripotent cells capa-ble of being reprogrammed in vitro to become functionalinsulin-producing cells.

Several in vivo studies (28–30) demonstrate that BMcells contribute to pancreatic �-cell regeneration at a lowfrequency, ranging from 1 to 2.7%. Ianus et al. (29) pro-vided in vivo evidence of adult mouse BM harboring cellsthat could transdifferentiate into glucose-competent pan-creatic endocrine cells using a cAMP response element–LoxP system, as assessed in cross-sex BM transplantexperiments. The above findings indicate that this in vivoprocess is likely due to transdifferentiation of BM-derivedcells into insulin-producing �-cells, rather cell fusion, asthe main source of BM-derived hepatocytes repopulatingthe liver of the mice, as was the case with fumarylaceto-acetate hydrolase deficiency, suggested by other studies(15,31). In our current study, the homogeneous mBMDScells were used to induce in vitro differentiation intoinsulin-producing cells; hence, cell fusion is likely not theanswer for the presence of competent insulin-producingcells. Other studies confirm that allogeneic BM transplan-tation with as low as 1% chimerism in pancreatic islets canreverse the diabetogenic process in pre-diabetic mice (28).Hess et al. (30) have provided a theory based on theirobservations in a recent study to explain the possiblemechanism of reversing hyperglycemia in diabetic mice

after BM transplantation. They suggest that pancreaticengraftment of donor BM–derived cells expressing endothe-lial markers after BM transplantation initiate endogenous�-cell regeneration, whereas donor BM–derived insulin-pos-itive �-cells represent a rare event. Kojima et al. (32) recentlyfound extrapancreatic proinsulin-producing cells present inthe liver, BM, spleen, adipose tissue, and thymus in hyper-glycemic animals and that the majority of these proinsulin-producing cells were derived from the donor BM, asevidenced by BM transplantation experiments. These studiessupport our observation that BM contains stem cells capableof differentiation into insulin-producing cells.

One of the key questions to address is which cell type inthe BM is responsible for pancreatic endocrine transdif-ferentiation. Unfortunately, it is difficult to extrapolate thecell phenotype from currently available studies using an invivo approach. Our results from both humans and micesuggest that CD45-negative adherent pluripotent mesen-chymal stem cells are capable of transdifferentiation intoinsulin-producing cells in vitro under high-glucose cultureconditions. The common phenotype of the BMDS cellsbetween humans and mice is CD45 negative, CD34 nega-tive, and C-kit negative, indicating that they are unlikely tobe hematopoietic stem cells. However, the possibility ofcirculating pancreatic stem cells cannot be completelyexcluded. A recent study published by Kodama et al. (33)indicated that injection of splenocytes into pre-diabeticNOD mice reversed diabetes and promoted pancreatic�-cell regeneration. Their results further indicated thatCD45-negative splenocytes (presumably mesenchymalprecursor cells) were responsible for islet �-cell regener-ation. This result indirectly supports our conclusion thatthe BM-derived stem cells capable of generating isletprecursor cells have an immunophenotype and biologiccharacteristics similar to those of BM mesenchymal cells.

There are two key steps in our cell culture conditionsthat appear important for inducing the differentiation ofBMDS cells into insulin-producing islet-like cells. First, themBMDS cells initially require culture in medium contain-ing a high-glucose concentration (23 mmol/l) for variousdurations of time until certain genes, such as Pdx-1, insulinI and II, Glut-2, and islet amyloid polypeptide, become

FIG. 7. A: Cell transplantation using D-mBMDS cells. Balb/c mice became diabetic within 12 days after two intraperitoneal injections of 250 and50 �g/g body wt STZ over 3 days apart. Glucose levels were monitored by tapped tail-vein blood at 1600 under nonfasting conditions. f, bloodglucose levels in sham surgery diabetic mice (n � 5); �, mBMDS cell–implanted mice after their glucose levels had reached >350 mg/dl (n � 6).The arrow indicates the day of implantation (shown as day 0). Values are means � SD. B: Glucose responses during the IPGT test. Glucosetolerance was tested following an intraperitoneal injection of glucose (2 mg/g body wt) in overnight-fasted control (Œ) (n � 5) and implanted (F)(n � 3) mice 14 days after transplantation. The venous blood was collected from the tail vein at 0, 30, 60, 90, 120, and 150 min after the injection.Values are means � SD.



detectable. Second, in order for the D-mBMDS cells tobecome glucose responsive, further differentiation andmaturation are required through either in vitro culturewith �-cell–promoting factors, such as nicotinamide andexendin 4, or transplantation of the cells into diabeticanimals. In this study, we demonstrated that mBMDS cellscultured under low-glucose conditions did not express theaforementioned genes, and in addition, they did not se-crete insulin upon glucose stimulation, even in the 7-daypresence of the �-cell–stimulating factors exendin 4 andnicotinamide. Our data indicate that long-term culture in ahigh-glucose medium reprogrammed these cells toward apathway of pancreatic endocrine cell differentiation. At acertain period of time (2–4 months) and via a still unclearmechanism, this switch occurred.

It is well known that glucose is a growth factor for�-cells (34). It promotes �-cell replication in vitro and invivo at a 20- to 30-mmol/l concentration (35) and increasesinsulin content in cell lines derived from embryonic stemcells (20) at a 5-mmol/l concentration. The effect ofchronic hyperglycemia on pancreatic �-cells, however,remains controversial. In an in vivo study, Jonas et al. (36)

showed that the expression of several genes important forglucose-stimulated insulin secretion (glucose metabolismenzymes and ion channels/pumps) was gradually de-creased with increasing levels of blood glucose. They alsosuggested a link between stimulation of �-cell growth anda reduced state of differentiation in hyperglycemic ani-mals. However, these observations were primarily focusedon pancreatic �-cells. The effects of long-term high-glu-cose culture on stem cells (adult or embryonic) wereunclear. In a previous study, we demonstrated that along-term culture of purified hepatic oval stem cells inhigh-glucose (23 mmol/l) medium promoted the oval cellsto transdifferentiate into functional insulin-secreting cells(16). In addition, we have observed that overexpression ofPdx-1 in a hepatic stem cell line (WB cells) only results inthe generation of pancreatic precursor cells (unpublishedobservations). These cells did not respond to a glucosechallenge in vitro by releasing insulin. Rather, these pre-cursor cells became fully functional under two conditions:one involving long-term culture in high-glucose mediumand the other being the transplantation of these cells intodiabetic (i.e., hyperglycemic) animals.

FIG. 8. Flow cytometric analysis of apoptotic cells. Cells were cultured under high-glucose conditions for the indicated time periods, thencollected for flow analysis. Intact cells were stained with Annexin-PE to detect outer-leaf phosphoserine, which is indicative of bilayer flippingand membrane blebbing in early apoptosis, and counterstained with the intercalating dye, 7AAD, to detect membrane permeability and chromatindegradation. Cells that stained with Annexin V but did not take up the 7AAD (upper left quadrant of each contour plot) were considered“apoptotic.” Cells that had background levels of Annexin V staining (set on untreated cell cultures, upper left panel) and no uptake of 7AAD wereconsidered “viable and nonapoptotic.” The percentage of total cells (out of 10,000 detected events) in the “apoptotic” (upper left) quadrant andthe “viable” (lower left) quadrant are given. PE, phycoerythrin



The notion that in vitro high-glucose culture (or in vivohyperglycemia) represents a critical factor for adult stemcell transdifferentiation into insulin-producing cells hasbeen supported by recent two publications. Zalzman et al.(37) demonstrated that culture of immortalized humanfetal Pdx-1–expressing hepatocytes in media containing 25mmol/l glucose activated multiple �-cell genes, producedand stored considerable amounts of insulin, and releasedinsulin in a regulated manner. In another work, Kojima etal. (32) showed that it was hyperglycemia produced by a25% glucose injection into nondiabetic mice as well as inthree other types of hyperglycemic animal models that ledto the appearance of proinsulin-positive cells within 3 daysin the liver, fat, spleen, BM, and thymus, as well asinsulin-positive cells within 15 days in those organs. Thesestudies support our observation that both liver and BM-containing stem cells can be induced under high-glucoseconditions to differentiate into insulin-producing cells, andthat insulin-producing cells can be derived from liver andBM cells. A sharp difference between our in vitro obser-vations (in months) and those of the in vivo works of Kojimaet al. (in days) (32) most likely resides in the requiredduration of exposure in terms of the need for high-glucoseconditions to generate insulin-producing cells. Possible ex-planations may include that 1) in vivo three-dimensionalstructure and cell-cell contact and interaction may play avital role in promoting cell differentiation, 2) other solublefactors in addition to high glucose in vivo may also play a rolein accelerating cell differentiation, and 3) it takes a long timefor a single-cell–derived cell clone to form three-dimensionalcell clusters under high-glucose culture conditions. Theabove theory is supported by our observations involving ourdetection of an increase in insulin in culture medium takenfrom 3-week cultures of whole-marrow adherent cells underhigh-glucose conditions.

One possible explanation of the constant viable cells inthe high-glucose cultures is that cycle events occur be-tween immature stem-like cells and D-mBMDS cells. Theimmature stem-like cells, which may be more resistant tohigh-glucose culture conditions, give rise to the D-mBMDScells, which may be more susceptible to high-glucose-induced apoptosis. These early apoptotic cells were re-moved following either medium change or cell splits. Thisassumption helps us explain the strange behavior of thecells we have repeatedly observed, namely the loss ofpancreatic endocrine gene expression and insulin produc-tion when these cells were quickly expanded to a largequantity for cell transplantation experiments. It has beenproposed that rapid growth will reduce cell differentiation.In our system, rapid proliferation may be only a part of thestory for the loss of cell differentiation, the other facetbeing related to the high-glucose–induced cell apoptosis ofthe D-mBMDS cells and the subsequent loss of the differ-entiated cells. The latter may explain the variations be-tween experiments and the loss in expression of genesrelated to pancreatic endocrine differentiation. Second,our experience with transdifferentiation of the humanBMDS cells also indicated the requirement for subsequentculture with maturation factors, such as exendin-4 andnicotinamide, in a medium containing low FCS and lowglucose to promote cell maturation and to restore thesensitivity to a glucose challenge (D.-Q.T., L.-Z.C., S.A.L.,

M.A.A., L.-J.Y., unpublished results). GLP-1 is an incretinhormone capable of restoring normal glucose tolerance inaging glucose-intolerant Wistar rats and inducing differen-tiation of islet Pdx-1–positive ductal cells into insulin-secreting cells (38). GLP-1 stimulates insulin secretion andaugments �-cell mass via activation of �-cell proliferationand islet neogenesis (13). A recent study by Suzuki et al.(25) demonstrated that GLP-1 converts intestinal epithelialcells into functional insulin-producing cells. Exendin-4 is apotent GLP-1 agonist that has previously been shown tostimulate both �-cell replication and neogenesis fromductal progenitor cells (39). We have shown that the latestage of D-mBMDS cells expressed the GLP-1 receptorgene and that this expression may correlate with glucose-responsive insulin release.

Nicotinamide is a poly(ADP-ribose) synthetase inhibitorknown to differentiate and increase cell mass in culturedhuman fetal pancreatic cells (40) and to protect cells fromdesensitization induced by prolonged exposure to largeamounts of glucose. Sjoholm, Korsgren, and Andersson(41) demonstrated that nicotinamide promoted formationof fetal porcine islet-like cell clusters and increased therates of proinsulin biosynthesis in these clusters. Theyconcluded that the stimulatory effects of nicotinamide oninsulin production and content by fetal porcine islet-likecell clusters result from neoformation of �-cells throughdifferentiation. Finally, a report by Ramiya et al. (24)described how nicotinamide-treated islets derived fromthe pancreatic progenitor cell had more insulin and se-creted significantly more insulin than cultures treated withglucose alone. Our previously published study (16)showed that nicotinamide promotes in vitro transdifferen-tiation and maturation of the liver stem cells into insulin-producing cells. Taken together, a combination of exendin4 and nicotinamide effectively promotes further D-mBMDScell differentiation in our experimental system.

Our present study demonstrates the potential for cell-based therapy of diabetes involving the generation ofautologous insulin-producing cells in vitro from BMDScells. These in vitro–generated insulin-producing cellscould, in theory, provide a potentially unlimited source ofislet-like cells without the limitation of immune rejectionbased on alloimmunity. However, because there are mul-tifactorial influences in the transdifferentiation of BM-derived stem cells into competent insulin-producing cells,there are many questions left unanswered and unresolvedissues remain. Among those are answers to the questionsof what are the decisive steps (e.g., addition of the glucose,exogenous factors, and timing of factor addition) for thetransdifferentiation process to take place? In our experi-ence, these differentiated cells are unlike �-cell–derivedcell lines, such as �-TC and INS-1 cells, in terms of theirgene expression profiles, cell maturity, and capacity torelease insulin in response of glucose stimulation (data notshown). Hence, one can also question whether these cellscan really be pushed to the level of maturity like true�-cells by changing the in vitro culture conditions. Anotherrelevant clinical question involves the issue of autoimmu-nity. Will the immune response to �-cell antigens recognizeand destroy the newly generated insulin-producing cellsobtained from BMDS cells? Obviously, further research isrequired to address these important questions. Yet we believe



the results demonstrated in this study provide direct evi-dence supporting the notion that transdifferentiation of adultstem cells to insulin-producing cells may represent a viabletherapeutic option for type 1 diabetes.

ACKNOWLEDGMENTS

This work was supported by grants R21-DK063270 (toL.-J.Y.) and K08-DK064054 (to L.-J.Y.) from the NationalInstitutes of Health and by the Juvenile Diabetes ResearchFoundation.

We thank Dr. Jill Verlander Reed and Kim Ahren fortechnical assistance.

REFERENCES

1. Atkinson MA, Eisenbarth GS: Type 1 diabetes: new perspectives on diseasepathogenesis and treatment. Lancet 358:221–229, 2001

2. Shapiro AM, Lakey JR, Ryan EA, Korbutt GS, Toth E, Warnock GL,Kneteman NM, Rajotte RV: Islet transplantation in seven patients with type1 diabetes mellitus using a glucocorticoid-free immunosuppressive regi-men. N Engl J Med 343:230–238, 2000

3. Gunnarsson R, Klintmalm G, Lundgren G, Wilczek H, Ostman J, Groth CG:Deterioration in glucose metabolism in pancreatic transplant recipientsgiven cyclosporin. Lancet 2:571–572, 1983

4. Kopen GC, Prockop DJ, Phinney DG: Marrow stromal cells migratethroughout forebrain and cerebellum, and they differentiate into astrocytesafter injection into neonatal mouse brains. Proc Natl Acad Sci U S A

96:10711–10716, 19995. Asahara T, Murohara T, Sullivan A, Silver M, van der ZR, Li T, Witzen-

bichler B, Schatteman G, Isner JM: Isolation of putative progenitorendothelial cells for angiogenesis. Science 275:964–967, 1997

6. Asahara T, Masuda H, Takahashi T, Kalka C, Pastore C, Silver M, Kearne M,Magner M, Isner JM: Bone marrow origin of endothelial progenitor cellsresponsible for postnatal vasculogenesis in physiological and pathologicalneovascularization. Circ Res 85:221–228, 1999

7. Ferrari G, Cusella-De Angelis G, Coletta M, Paolucci E, Stornaiuolo A,Cossu G, Mavilio F: Muscle regeneration by bone marrow-derived myo-genic progenitors. Science 279:1528–1530, 1998

8. Gussoni E, Soneoka Y, Strickland CD, Buzney EA, Khan MK, Flint AF,Kunkel LM, Mulligan RC: Dystrophin expression in the mdx mouserestored by stem cell transplantation. Nature 401:390–394, 1999

9. Pereira RF, Halford KW, O’Hara MD, Leeper DB, Sokolov BP, Pollard MD,Bagasra O, Prockop DJ: Cultured adherent cells from marrow can serve aslong-lasting precursor cells for bone, cartilage, and lung in irradiated mice.Proc Natl Acad Sci U S A 92:4857–4861, 1995

10. Schwartz RE, Reyes M, Koodie L, Jiang Y, Blackstad M, Lund T, Lenvik T,Johnson S, Hu WS, Verfaillie CM: Multipotent adult progenitor cells frombone marrow differentiate into functional hepatocyte-like cells. J Clin

Invest 109:1291–1302, 200211. Petersen BE, Bowen WC, Patrene KD, Mars WM, Sullivan AK, Murase N,

Boggs SS, Greenberger JS, Goff JP: Bone marrow as a potential source ofhepatic oval cells. Science 284:1168–1170, 1999

12. Theise ND, Badve S, Saxena R, Henegariu O, Sell S, Crawford JM, KrauseDS: Derivation of hepatocytes from bone marrow cells in mice afterradiation-induced myeloablation. Hepatology 31:235–240, 2000

13. Li Y, Hansotia T, Yusta B, Ris F, Halban PA, Drucker DJ: Glucagon-likepeptide-1 receptor signaling modulates beta cell apoptosis. J Biol Chem

278:471–478, 200314. Theise ND, Nimmakayalu M, Gardner R, Illei PB, Morgan G, Teperman L,

Henegariu O, Krause DS: Liver from bone marrow in humans. Hepatology

32:11–16, 200015. Vassilopoulos G, Wang PR, Russell DW: Transplanted bone marrow

regenerates liver by cell fusion. Nature 422:901–904, 200316. Yang L, Li S, Hatch H, Ahrens K, Cornelius JG, Petersen BE, Peck AB: In

vitro trans-differentiation of adult hepatic stem cells into pancreaticendocrine hormone-producing cells. Proc Natl Acad Sci U S A 99:8078–8083, 2002

17. Ohgawara H, Kawamura M, Honda M, Karibe S, Iwasaki N, Tasaka Y,Omori Y: Reversal of glucose insensitivity of pancreatic B-cells due toprolonged exposure to high glucose in culture: effect of nicotinamide onpancreatic B-cells. Tohoku J Exp Med 169:159–166, 1993

18. Li SW, Tang D, Ahrens KP, She JX, Braylan RC, Yang L: All-trans-retinoic

acid induces CD52 expression in acute promyelocytic leukemia. Blood

101:1977–1980, 200319. Hohmeier HE, Mulder H, Chen G, Henkel-Rieger R, Prentki M, Newgard

CB: Isolation of INS-1–derived cell lines with robust ATP-sensitive K�

channel–dependent and –independent glucose-stimulated insulin secre-tion. Diabetes 49:424–430, 2000

20. Soria B, Roche E, Berna G, Leon-Quinto T, Reig JA, Martin F: Insulin-secreting cells derived from embryonic stem cells normalize glycemia instreptozotocin-induced diabetic mice. Diabetes 49:157–162, 2000

21. Elliott JI, Dewchand H, Altmann DM: Streptozotocin-induced diabetes inmice lacking alphabeta T cells. Clin Exp Immunol 109:116–120, 1997

22. Jiang Y, Jahagirdar BN, Reinhardt RL, Schwartz RE, Keene CD, Ortiz-GonzalezXR, Reyes M, Lenvik T, Lund T, Blackstad M, Du J, Aldrich S, Lisberg A, LowWC, Largaespada DA, Verfaillie CM: Pluripotency of mesenchymal stem cellsderived from adult marrow. Nature 418:41–49, 2002

23. Bonner-Weir S, Taneja M, Weir GC, Tatarkiewicz K, Song KH, Sharma A,O’Neil JJ: In vitro cultivation of human islets from expanded ductal tissue.Proc Natl Acad Sci U S A 97:7999–8004, 2000

24. Ramiya VK, Maraist M, Arfors KE, Schatz DA, Peck AB, Cornelius JG:Reversal of insulin-dependent diabetes using islets generated in vitro frompancreatic stem cells. Nat Med 6:278–282, 2000

25. Suzuki A, Nakauchi H, Taniguchi H: Glucagon-like peptide 1 (1-37)converts intestinal epithelial cells into insulin-producing cells. Proc Natl

Acad Sci U S A 100:5034–5039, 200326. Lumelsky N, Blondel O, Laeng P, Velasco I, Ravin R, McKay R: Differen-

tiation of embryonic stem cells to insulin-secreting structures similar topancreatic islets. Science 292:1389–1394, 2001

27. Assady S, Maor G, Amit M, Itskovitz-Eldor J, Skorecki KL, Tzukerman M:Insulin production by human embryonic stem cells. Diabetes 50:1691–1697, 2001

28. Zorina TD, Subbotin VM, Bertera S, Alexander AM, Haluszczak C, Gambrell B,Bottino R, Styche AJ, Trucco M: Recovery of the endogenous beta cellfunction in the NOD model of autoimmune diabetes. Stem Cells 21:377–388,2003

29. Ianus A, Holz GG, Theise ND, Hussain MA: In vivo derivation of glucose-competent pancreatic endocrine cells from bone marrow without evidenceof cell fusion. J Clin Invest 111:843–850, 2003

30. Hess D, Li L, Martin M, Sakano S, Hill D, Strutt B, Thyssen S, Gray DA,Bhatia M: Bone marrow-derived stem cells initiate pancreatic regenera-tion. Nat Biotechnol 21:763–770, 2003

31. Wang X, Willenbring H, Akkari Y, Torimaru Y, Foster M, Al Dhalimy M,Lagasse E, Finegold M, Olson S, Grompe M: Cell fusion is the principalsource of bone-marrow-derived hepatocytes. Nature 422:897–901, 2003

32. Kojima H, Fujimiya M, Matsumura K, Nakahara T, Hara M, Chan L:Extrapancreatic insulin-producing cells in multiple organs in diabetes.Proc Natl Acad Sci U S A 101:2458–2463, 2004

33. Kodama S, Kuhtreiber W, Fujimura S, Dale EA, Faustman DL: Isletregeneration during the reversal of autoimmune diabetes in NOD mice.Science 302:1223–1227, 2003

34. Soria B: In-vitro differentiation of pancreatic beta-cells. Differentiation

68:205–219, 200135. Bonner-Weir S, Deery D, Leahy JL, Weir GC: Compensatory growth of

pancreatic �-cells in adult rats after short-term glucose infusion. Diabetes

38:49–53, 198936. Jonas JC, Sharma A, Hasenkamp W, Ilkova H, Patane G, Laybutt R,

Bonner-Weir S, Weir GC: Chronic hyperglycemia triggers loss of pancreaticbeta cell differentiation in an animal model of diabetes. J Biol Chem

274:14112–14121, 199937. Zalzman M, Gupta S, Giri RK, Berkovich I, Sappal BS, Karnieli O, Zern MA,

Fleischer N, Efrat S: Reversal of hyperglycemia in mice by using humanexpandable insulin-producing cells differentiated from fetal liver progeni-tor cells. Proc Natl Acad Sci U S A 100:7253–7258, 2003

38. Hui H, Wright C, Perfetti R: Glucagon-like peptide 1 induces differentiationof islet duodenal homeobox-1–positive pancreatic ductal cells into insulin-secreting cells. Diabetes 50:785–796, 2001

39. Xu G, Stoffers DA, Habener JF, Bonner-Weir S: Exendin-4 stimulates both�-cell replication and neogenesis, resulting in increased �-cell mass andimproved glucose tolerance in diabetic rats. Diabetes 48:2270–2276, 1999

40. Otonkoski T, Beattie GM, Mally MI, Ricordi C, Hayek A: Nicotinamide is apotent inducer of endocrine differentiation in cultured human fetal pan-creatic cells. J Clin Invest 92:1459–1466, 1993

41. Sjoholm A, Korsgren O, Andersson A: Polyamine requirements in nicoti-namide-stimulated beta-cell differentiation in fetal porcine islet-like cellclusters. Endocrinology 135:1559–1565, 1994



Documents

In Vivo and In Vitro Characterization of Insulin-Producing ......Cytospin slides from differentiated mBMDS (D-mBMDS) cells were made for insulin and C-peptide protein expression. The