Experimental Neurology 164, 247–256 (2000)doi:10.1006/exnr.2000.7389, available online at http://www.idealibrary.com on

Adult Bone Marrow Stromal Cells Differentiate into Neural Cells in VitroJ. Sanchez-Ramos,*,† S. Song,*,† F. Cardozo-Pelaez,*,† C. Hazzi,†,‡ T. Stedeford,† A. Willing,§

T. B. Freeman,§ S. Saporta,¶ W. Janssen,\ ,** N. Patel,†,†† D. R. Cooper,†,†† and P. R. Sanberg†,§*Department of Neurology, §Department of Neurosurgery, \Department of Medicine, ‡Department of Pharmacology,

¶Department of Anatomy, ††Department of Biochemistry and Molecular Biology, University of South Florida;**Moffitt Cancer Center; and †James Haley VA Hospital, Tampa, Florida

Received December 2, 1999; accepted February 4, 2000

Bone marrow stromal cells (BMSC) normally giverise to bone, cartilage, and mesenchymal cells. Re-cently, bone marrow cells have been shown to have thecapacity to differentiate into myocytes, hepatocytes,and glial cells. We now demonstrate that human andmouse BMSC can be induced to differentiate into neu-ral cells under experimental cell culture conditions.BMSC cultured in the presence of EGF or BDNF ex-pressed the protein and mRNA for nestin, a marker ofneural precursors. These cultures also expressed glialfibrillary acidic protein (GFAP) and neuron-specificnuclear protein (NeuN). When labeled human ormouse BMSC were cultured with rat fetal mesence-phalic or striatal cells, a small proportion of BMSC-derived cells differentiated into neuron-like cells ex-pressing NeuN and glial cells expressing GFAP.© 2000 Academic Press

Key Words: bone marrow stromal cells; stem cells;neuronal differentiation.

INTRODUCTION

The ancient Chinese believed bone marrow was thesource of brain tissue as suggested by the maxim“brain is a sea of marrow” (17). The existence of stemcells for nonhematopoietic cells in bone marrow wasproposed over 100 years ago, but the isolation anddifferentiation of marrow stromal cells into osteo-blasts, chondroblasts, adipocytes, and myoblasts wasonly recently demonstrated (see review (21)). Nonhe-matopoietic precursors from bone marrow stroma havebeen referred to as colony-forming-unit fibroblasts,mesenchymal stem cells, or bone marrow stromal cells(BMSC). Although BMSC can naturally be expected tobe a source of surrounding tissue of bone, cartilage, andfat, several recent reports demonstrate that these cells,under specific experimental conditions, can differenti-ate into muscle, glia, and hepatocytes (1, 8, 20). Bonemarrow cells also have the capacity to migrate exten-sively. Transplantation of genetically labeled bonemarrow cells into immunodeficient mice has been re-

247

ported to result in migration of marrow cells into aregion of chemically induced muscle degeneration (8).These marrow-derived cells underwent myogenic dif-ferentiation and participated in the regeneration of thedamaged muscle fibers. Systemic infusion of geneti-cally labeled bone marrow cells into irradiated femalemice resulted in an influx of labeled cells into the brainover days to weeks (6). Marrow-derived cells werefound throughout regions of the brain, from cortex tobrain stem. Some bone marrow-derived cells were pos-itive for the microglial antigenic marker F4/80. Othermarrow-derived cells expressed the astroglial markerglial fibrillary acidic protein (GFAP) (6). These resultsindicated that some microglia and astroglia arose froma precursor that is a normal constituent of adult bonemarrow. Other researchers have reported that infusionof human BMSC into rat striatum resulted in engraft-ing, migration, and survival of cells (1). After engraft-ment, these cells lost markers typical of marrow stro-mal cells in culture, such as immunoreactivity to anti-bodies against collagen and fibronectin. BMSCdeveloped many of the characteristics of astrocytes,and their engraftment and migration markedly con-trasted with fibroblasts that continued to produce col-lagen and undergo gliosis after implantation. Graftingof BMSC into the lateral ventricle of neonatal miceresulted in their migration throughout the forebrainand cerebellum without disruption of host brain archi-tecture (13). Some BMSC in striatum and hippocam-pus were reported to express GFAP. Moreover, occa-sional neurofilament-positive BMSC were found in thebrain stem suggesting that some BMSC differentiatedinto a neuronal phenotype (13). All of these reportsprovide impetus to investigate the potential of bonemarrow cells to develop into nonhematopoietic cellsand, in particular, to generate neural lineages. To testthe hypothesis that neural precursor cells can be de-rived from bone marrow cells, stromal cells from themarrow were induced to proliferate and then culturedunder conditions that induce neuronal differentiationof embryonic or neural stem cells in vitro (5, 24).

0014-4886/00 $35.00Copyright © 2000 by Academic Press

All rights of reproduction in any form reserved.

raw

l

ri

pti

248 SANCHEZ-RAMOS ET AL.

METHODS

Animals

C57BL/6J or transgenic LacZ mice (C57BL/6J-Gtrosa26) were purchased from JAX Labs, Inc. Timed-pregnant rats (Sprague–Dawley) were purchased fromHarlan, Inc.

Preparation of Mouse Bone Marrow Cells

Mouse bone marrow cells were collected, after sacri-fice of 8-week-old mice, from femurs and tibias byflushing the shaft with buffer (phosphate-buffered sa-line supplemented with 0.5% bovine serum albumin,pH 7.2) using a syringe with a No. 26 G needle. Cellswere disaggregated by gentle pipetting several times.Cells were passed through 30-mm nylon mesh to re-move remaining clumps of tissue. Cells were washedby adding buffer, centrifuging for 10 min at 200g andremoving supernatant. The cell pellet was resuspendedin 800 ml of buffer for each 108 cells.

Separation of Mouse Bone Marrow HematopoieticStem Cell Antigen (Sca11) Cells

Using a magnetic cell sorting kit (Milteny Biotec,Inc., Auburn, TX), bone marrow cells were labeled withSca11 or CD341 microbeads, which label bone mar-ow cells that express stem cell antigen (Sca1 in mousend CD34 in human). The labeled bone marrow cellsere passed through an MS1 column for positive se-

lection of Sca11 or CD341 cells. Two hundred micro-iters of Sca1 Multi-sort MicroBeads was added per 108

total cells, mixed, and incubated for 15 min at 6–12°C.Cells were washed by adding 5–10X the labeling vol-ume of buffer, centrifuging for 10 min at 200g, andemoving supernatant. The cell pellet was resuspendedn 500 ml buffer. The MS1/RS1 column was washed

with 500 ml of buffer. The cell suspension was appliedto the column and the negative cells were passedthrough. The column was then rinsed with 500 ml ofbuffer three times. The column was removed from theseparator (which contains the magnet) and placed on asuitable collection tube. One milliliter of buffer waspipetted onto the column and the positive fraction wasflushed out with the plunger provided with the column.

Separation of BMSC from Human Bone MarrowAspirates

Human bone marrow stromal cells were harvestedfrom residual material (bone chips with adherent stro-mal cells, fatty tissue, and debris) retained on thenylon filters routinely used to clean freshly procuredhuman marrow aspirates. The filtrate, which was notused in the present studies, was destined for bonemarrow replacement therapies and contains the bulk

of bone marrow cells, including the hematopoietic pre-cursor cells. The nylon filter contained material that isusually discarded but was utilized as the starting ma-terial for generation of neural cells. This filter was backwashed five times with normal sterile saline and cen-trifuged to remove bone chips. The bone marrow ma-terial was diluted 1:1 with Dulbecco’s minimal essen-tial media (DMEM, GIBCO/BRL) and 10% fetal bovineserum (FBS) and centrifuged through a density gradi-ent (Ficoll–Paque Plus, 1.077 g/ml, Pharmacia) for 30min at 1000 g. The supernatant and interface werecombined and diluted to approximately 20 ml withgrowth medium and plated in polyethylene-imine-coated plastic flasks. The growth medium consisted ofDMEM supplemented with 2 mM glutamine, 0.001%b-mercaptoethanol, nonessential amino acids, 10%horse serum, and either human leukemia inhibitoryfactor (hLIF) or epidermal growth factor (EGF), 10ng/ml. The cells were incubated at 37°C in 5% CO2 inflasks for 2 days and nonadherent cells were removedby replacing the medium. After the cultures reachedconfluency, usually within a week, the cells were liftedby incubation with trypsin (0.25%) and 1 mM EDTA at37°C for 3–4 min. They were then frozen for later useor replated after 1:2 or 1:3 dilution with the addition ofEGF (10 ng/ml). Cells used in these experiments wereharvested from the first or second passage.

Preparation of Rat or Mouse Fetal Midbrain CellSuspensions for Coculturing with BMSC Cells

Fetal midbrain cell suspensions were prepared fromventral midbrain from rat or mouse embryos at 14 daysof gestation dissected in HBSS (Hanks’ balanced saltsolution; GIBCO BRL, Life Technologies, Gaithers-burg, MD) 115 mM Hepes (GIBCO BRL) as previouslydescribed (15, 23). After 3 days in vitro, BMSC were

FIG. 1. BMSC adherent to culture dishes were treated with EGF(10 ng/ml), RA (0.5 mM), or RA plus BDNF (10 ng/ml) for 7 days. Eachbar represents the mean number (6SEM) of fibronectin-immunore-active cells per visual field (203 objective) determined in 20 fields perdish in four culture dishes. *P , 0.05, using two-tailed t test com-

aring number of fibronectin-ir cells treated with RA 1 BDNF tohose treated with EGF or RA. The decrease in total number of cellss not significantly different under each condition.

249BONE MARROW CELLS DIFFERENTIATE INTO NEURAL CELLS

added to the midbrain culture layer in an approxi-mately 1:1 ratio of cells.

Differentiation Procedure

Cells removed from the flask bottom after the first orsecond passage as above (BMSC) were replated in35-mm culture dishes in the presence of a neuronalgrowth medium (N5) (11) supplemented with 5% horseserum, 1% FBS, transferrin (100 mg/ml), putrescine (60mM), insulin (25 mg/ml), progesterone (0.02 mM), sele-nium (0.03 mM), all-trans-retinoic acid (0.5 mM), andBDNF at a concentration of 10 ng/ml. After 7 to 14days, BMSC were processed for immunocytochemistry,Western blotting, or rt-PCR.

FIG. 2. Human BMSC grown (A) in the presence of EGF and (Bimmunoreactivity. Notice smaller, ovoid nonstained cells. Phase-con

Immunocytofluorescence and Immunocytochemistry

Bone marrow stromal cells were visualized followingfixation (4% buffered paraformaldehyde) and process-ing for immunocytochemistry with antibodies to fi-bronectin (rabbit polyclonal, Sigma, Inc.) (1:400) fol-lowed by biotinylated anti-rabbit antibody and an avi-din–biotin conjugate of horseradish peroxidase(Vectorstain, Vector Laboratory Burlingame, CA). Vi-sualization of neural cells (neurons and glia) utilizedimmunocytofluorescence technique and confocal fluo-rescence microscopy. In the coculture experiments, hu-man BMSC were prelabeled with 20 mM concentrationof fluorescent green “cell tracker” (5-chloromethyl fluo-

presence of RA 1 BDNF for 1 week. Brown stain shows fibronectinst photomicrograph. Scale bar 5 20 mm.

) intra

l

E

citccldctpbAv8oc

W

blHop

cnfi2v(c

nlr

(

250 SANCHEZ-RAMOS ET AL.

recein diacetate, Molecular Probes, Inc.) or red fluores-cent PKH-26 (Sigma, Inc.) and plated in 35-mm poly-ethyleneimine-coated culture dishes. After 2 or 3 days,BMSC cells were added to the rat midbrain cells. After1–2 weeks in culture, cells were processed for immu-nocytofluorescence using primary antibodies againstNeuN (mouse monoclonal, Chemicon, Inc.) (1:100),GFAP (rabbit polyclonal, Biogenex, Inc.) (1:100), nestin(mouse monoclonal, Chemicon, Inc.) (1:250), and MAP2(mouse monoclonal, Chemicon, Inc.) (1:200), followedby Texas red or fluorescein-labeled secondary antibody.

FIG. 3. Human BMSC were cultured for 7 days in all-trans-retinoic acid (0.5 mM) and BDNF (10 ng/ml). Fluorescence immuno-ytochemistry was used to detect the presence of (A) fibronectin, (B)estin, (C) neuron-specific nuclear protein (NeuN), and (D) glialbrillary acidic protein (GFAP). Frequency of fibronectin-ir cells '.2 cells per visual field or 22.5% of total cells; Nestin ' 1 per 30isual fields (0.5% of total cells); NeuN-ir ' 1 per 30 visual fields0.5% of total cells); GFAP-ir ' 1 per 20 visual fields (0.8% of totalells). Scale bar 5 20 mm.

FIG. 4. Western blots of cell culture lysates of human BMSCcultures grown under four conditions: (1) EGF (10 ng/ml) in prolif-eration medium; (2) N5 medium alone; (3) trans-retinoic acid (0.5mM) in N5 medium; and (4) trans-retinoic acid (0.5 mM), BDNF (10

g/ml) in N5 medium. Nestin immunoreactivity, relative to a-tubu-in, is greatly diminished in cultures treated with retinoic acid. Theseesults were consistent across three separate experiments.

A Zeiss laser scanning confocal microscope (ModelLSM 510) use used to visualize fluorescence in twoseparate channels. For work with BMSC obtained fromtransgenic lacZ mouse (Jackson Labs), b-gal stainingwas performed using a kit for light-microscopic visual-ization of the blue reaction product catalyzed by b-ga-actosidase (Invitrogen, Inc.).

stimates of Cell Number

Estimates of positively stained cells were based onounting cells in 20 random visual fields (20X objective)n four culture dishes for each marker in a minimum ofhree different experiments. When positively stainedells were present in every field (e.g., fibronectin-irells), the mean and SEM of cells per field were calcu-ated and compared with cell counts prepared underifferent conditions using a two-tailed t test. Whenells stained with antibodies to nestin, GFAP, or NeuN,heir presence was rare and estimates of the number ofositive cells was made by counting the average num-er of visual fields viewed before finding a positive cell.cell that was found on the average after viewing 20

isual fields (each field of which contains an average oftotal cells) corresponds to approximately 1/160 cells

r 0.65% of the total number of cells and 1 per 30 fieldsorresponds to 0.4% of total cells.

estern Blotting

Cultures were washed three times in cold phosphate-uffered saline (PBS), scraped into ice-cold PBS, andysed in an ice-cold lysis buffer containing 20 nM Tris–Cl (pH 8.0), 0.2 mM EDTA, 3% Nonidet P-40, 2 mM

rthovanadate, 50 mM NaF, 10 mM sodium pyrophos-hate, 100 mM NaCl, and 10 mg each of aprotinin and

leupeptin per milliliter. After incubation on ice for 10min, the samples were centrifuged at 14,000g for 15min and the supernatants were collected. An aliquotwas removed for total protein estimation (Bio-Rad as-say). An aliquot corresponding to 10 mg of total proteinof each sample was separated by SDS–PAGE (10%)under reducing conditions and transferred electro-phoretically to nitrocellulose filters. Nonspecific bind-ing of antibody was blocked with 5% nonfat dry milkovernight at 4°C. Immunoblotting was carried out with

FIG. 5. rt-PCR analysis of human bone marrow cells using nes-tin primers. b-Actin serves as an internal control. BMSC were grownin the presence of EGF (10 ng/ml) or retinoic acid (0.5 mM) 1 BDNF10 ng/ml). M, molecular marker.

Gwl

251BONE MARROW CELLS DIFFERENTIATE INTO NEURAL CELLS

rabbit polyclonal antibody to GFAP (Biogenex, Inc.) ormouse monoclonal antibody to nestin or neuron-spe-cific nuclear protein (NeuN) (both purchased fromChemicon, Inc.) followed by peroxidase-conjugated sec-ondary anti-immunoglobulin antibodies, and the blotswere developed by the enhanced chemiluminescencemethod (ECL, Amersham).

Reverse Transcriptase Polymerase Chain Reaction(rt-PCR)

Total RNA from cells and tissues was isolated by theguanidine isothiocyanate method (3) and total RNA(2–5 mg) was reverse transcribed using the Superscriptkit (GIBCO/BRL). The upstream sense primer for de-tection of human nestin was 59 CAGGCTTCTCTTG-

CTTTCTGG 39 and the downstream antisense primeras 59 TGGTGAGGGTTGAGGTTTGT 39. Probes were

abeled with [32P]dCTP using a random primer labelingsystem (Promega). After hybridization, membraneswere washed as described (9). Following 30 cycles ofamplification (94°C, 1 min; 56°C, 1 min; and 72°C, 1min) in a Hybaid thermoblock, 25% of the PCR reactionwas resolved on a 1.2% agarose gel. A 360-bp band wasobserved under UV light and photographed. Each anal-ysis was repeated three times with similar results.

RESULTS

Treatment of BMSC cultures with hLIF failed toinduce proliferation of mouse and human cells, regard-less of whether they were enriched or depleted of he-matopoietic stem cells. Instead, incubation of BMSCwith hLIF resulted in differentiation into cells withfibroblastic morphologies. This is very unlike the ef-fects of hLIF on mouse embryonic stem cells in whichthis factor has been shown to be essential for keepingembryonic stem cells in a proliferating, undifferenti-ated state (4, 25). The remainder of experiments de-scribed below utilized EGF (10 ng/ml) as the mitogen to

FIG. 6. Coculture of BMSC from lacZ mice with fetal midbrain cuand BDNF (10 ng/ml). (A) b-Gal1 cells (from lacZ BMSC). Cellphotmicrograph, objective 203). (B) Same visual field as in A, withactivity (NeuN-ir). Note that cell No. 1 is not lacZ-positive, but isNeuN-ir originate from BMSC.

induce proliferation of the plated BMSC. Other re-searchers have reported that the LIF family of cyto-kines does not promote proliferation of neural precur-sor cells in an undifferentiated state, similar to ourobservations of their effect on marrow cells (12). In-stead, they have been implicated in the regulation ofastrocyte differentiation from neural precursors of thedeveloping forebrain (12).

Differentiation of BMSC into Neuron-like andGlial-like Cells

Incubation of BMSC with “differentiation medium”that contained 10% fetal calf serum (FCS), 0.5 mMall-trans-retinoic acid (RA), and BDNF (10 ng/ml) re-sulted in a significant decrease in the proportion offibronectin-ir cells in the cultures (see Fig. 1). Therewas a trend for the total number of cells in the culturesto decrease but this did not reach statistical signifi-cance. The decrease in numbers of fibronectin-ir cellswas primarily due to loss of expression of fibronectinand a concomitant change in cell morphologies (Figs.2A and 2B). The large flat fibronectin-containingBMSC were transformed into spindle-shaped fibroblas-tic appearing cells with long processes. Many smalleroval cells with short processes did not show fibronectinimmunoreactivity. Immunocytofluorescence analysisof these cultures revealed the occasional presence ofnestin-ir, NeuN-ir, and GFAP-ir cells, which weremuch smaller in size than the fibronectin-ir cells (Fig.3). These cells were oval or spindle-shaped with shortprocesses and were present in very much smaller pro-portion than the predominant fibronectin-ir stromalcells. NeuN-ir and GFAP-ir cells numbered 0.5 and 1%,respectively, of the total number of fibronectin-ir cells.Western blots of cell culture lysates of these culturesdemonstrated the presence of the proteins nestin,NeuN, and GFAP in human BMSC (Fig. 4). The ex-pression of nestin protein relative to a-tubulin wasdecreased in cultures treated with RA or RA 1 BDNF.

res for 2 weeks in medium with with all-trans-retinoic acid (0.5 mM)re numbered to compare with identical field in B. (Bright-fields exhibiting fluorescent neuron-specific nuclear antigen immunore-uN-ir, demonstrating origin from fetal midbrain. Other numbered

ltus acellNe

252 SANCHEZ-RAMOS ET AL.

These results were consistent across three separateexperiments. Interestingly, both GFAP and NeuNwere expressed by BMSC under culture all conditions,including growth in the presence of N5 medium alone.Nestin mRNA was also detected by rt-PCR (19) in RNAprepared from human BMSC cultured in the presenceof EGF (10 ng/ml) or retinoic acid (0.5 mM) and BDNF(10 ng/ml) (Fig. 5). Although there appears to be dif-

FIG. 7. (A–I) Mouse BMSC were labeled with a red fluorescent mfor 1 week. Red fluorescence (A, D, G) shows BMSC cells. Greenneuron-specific nuclear protein (NeuN). Double-labeled cells, withdifferentiated into a neuron-like phenotype (C, F, I). (J–L) Mouse BMwith fetal midbrain cells for 2 weeks. (J) Cells of BMSC origin. MAPMAP2 neurons, but double-labeled MAP2-ir cells with mature neuroscanning confocal microscope (Zeiss LSM 510). Scale bars 5 20 mm.

fering intensities of signal on the blots, this method cannot be used to quantify mRNA.

Cocultures of BMSC (Labeled with b-Gal orFluorescent Markers) and Fetal Rat Midbrain Cells

To determine whether cellular environment wouldaugment neuronal differentiation of BMSC-derived

er (PKH-26). They were then cocultured with fetal midbrain culturesrescence (B, E, H) marks cells that express the neuronal markerreen fluorescent nucleus and red cytoplasm, represent BMSC that

were labeled with red fluorescent marker (PKH-26) and coculturedmunoreactive neurons in green fluorescence. (L) BMSC adjacent tomorphology were not found. All images were obtained with a laser

arkfluoa gSC

2-imnal

nca

c

c

t

253BONE MARROW CELLS DIFFERENTIATE INTO NEURAL CELLS

cells, we studied the effects of coculturing BMSC withfetal primary neuronal cultures. The BMSC were iden-tified by using b-galactosidase-expressing bone mar-row cells prepared from transgenic lacZ mice. In asecond set of experiments, the BMSC were labeled witheither green fluorescent “cell tracker” or the red fluo-rescent PKH-26. The lacZ BMSC were plated (2.5 3105 cells/dish) on fetal mouse mesencephalic cell cul-tures (2.5 3 105 cells per dish) prepared as previouslydescribed (23). Retinoic acid (0.5 mM) and BDNF (10

g/ml) were added to the N5 medium. After 2 weeks,ultures were fixed and processed for histochemicalnd immunocytochemical staining. b-Galactosidase-

positive cells were clearly identified by blue stainingvisualized under bright field microscopy (See Fig. 6).

FIG. 8. (A–C) Mouse BMSC were labeled with red fluorescentfluorescence (A) illustrates cells of BMSC origin. Green fluorescencethat express GFAP (C). Scale bar 5 50 mm. (D–F) Mouse BMSC wefetal midbrain cultures for 1 week. Red fluorescence illustrates BMillustrates cells of bone marrow origin (E). Yellow fluorescence illust(G–I) Human BMSC were labeled with green cell tracker and then cmarks GFAP-ir cells (G). Green fluorescence marks cells of BMSC o20 mm. All images were obtained with a laser scanning confocal mic

Neurons were identified by NeuN immunoreactivity.Approximately 2–5% of b-galactosidase-positive cellswere also positive for NeuN, a much greater proportionof neuron-like cells than observed when BMSC werecultured alone. A very small proportion (,0.1%) ofontrol fetal midbrain cultures also stained positive for

b-gal, reflecting the endogenous expression of b-galac-tosidase activity (10). The increase in the number ofNeuN-ir BMSC in the presence of fetal neuronal cul-tures was significantly greater (at least twofold) thanwhen BMSC were cultured alone or when compared tonumbers of b-gal1 cells seen in control fetal midbrainultures.An alternative marker was also utilized to label

he BMSC. Human BMSC were labeled with 20 mM

and then cocultured with fetal midbrain cultures for 1 week. Redmarks GFAP-ir cells. Double-labeled cells (yellow) represent BMSCabeled with fluorescent green cell tracker and then cocultured withthat express the astroglial marker GFAP (D). Green fluorescences double-stained BMSC that express GFAP (F). Scale bar 5 20 mm.ltured with fetal rat midbrain cultures for 1 week. Red fluorescencen (H) and yellow cells show BMSC that express GFAP. Scale bar 5cope (Zeiss LSM 510).

dye(B)re lSC

rateocurigiros

ablsBedpttpGce

GfiBsewrpteb2ds

254 SANCHEZ-RAMOS ET AL.

red fluorescent PKH-26 (Sigma, Inc.) or fluorescentgreen “cell tracker” (5 chloromethyl fluorecein diac-etate, Molecular Probes, Inc.) and plated on a cellbed of rat midbrain cells prepared 3 days earlier.Cultures were fed with all-trans-RA (0.5 mM) 1BDNF (10 ng/ml) in N5 medium. After 1 week inculture, cells were processed for immunocytofluores-cence using primary antibodies against NeuN,GFAP, nestin, and MAP2, followed by Texas red orfluorescein-labeled secondary antibody. This permit-ted a dual labeling of cells to determine whether thecells that exhibited specific markers for neurons,glia, or nestin had a bone marrow origin (see Fig. 7).In cultures stained for NeuN immunoreactivity, 2 to4% of the BMSC-derived cells in culture were double-labeled and had the phenotype of immature neuron-like cells (Figs. 7A–7I). The cells were ovoid or spin-dle-shaped with bipolar processes. Labeled BMSCwere found adjacent to well-developed MAP2-ir neu-rons of fetal midbrain origin (Figs. 7J–7L), but noBMSC-derived cells were found that reliably ex-pressed MAP2 in a clearly neuronal phenotype.BMSC raised alone in the absence of fetal cell cul-tures were also found to not express MAP2. In cul-tures stained for GFAP immunoreactivity 5 to 8% ofthe BMSC-derived cells were double-labeled (Fig. 8).The majority of the double-labeled GFAP-ir cells hadan ovoid or spindle-shaped morphology with shortprocesses similar to those of the NeuN-ir cells (Fig.8I), but some GFAP-ir BMSC were irregular in shape(Fig. 8F) and approximated the morphology of glialcells.

DISCUSSION

These data provide evidence that human and murineBMSC are capable of differentiating into cells thatexpress several neural proteins and resemble imma-ture neurons or glial cells. Prior to the induction ofdifferentiation, the BMSC cultures were enriched infibronectin-ir cells and depleted of mouse hematopoi-etic stem cells (Sca1) or human hematopoietic stemcells (CD341). Treatment of the cultures with retinoiccid and BDNF resulted in decreased numbers of fi-ronectin-ir cells. This was due to gradual loss of thearge, flat fibronectin-ir cells and the appearance ofmaller ovoid or spindle-shaped cells. Analysis ofMSC lysates, prepared from cultures treated withither proliferation medium or differentiation medium,emonstrated the presence of nestin, NeuN, and GFAProtein. Treatment with RA or RA 1 BDNF decreasedhe expression of nestin protein. Microscopic examina-ion of the cultures, following immunocytochemicalrocessing, revealed a small proportion of NeuN-ir andFAP-ir cells (0.5 and 1%, respectively, of the BMSC

ells). Coculturing BMSC with lacZ fetal mouse mes-ncepahlic cells increased the number of NeuN-ir and

FAP-ir cells at least twofold. These results were con-rmed in a second set of experiments that utilizedMSC labeled with two different fluorescent vitaltains. The coculture experiments support the hypoth-sis that cell to cell contact, in addition to signalingith trophic factors and cytokines, plays an important

ole in differentiation of these BMSC. The neural cellsroduced from BMSC in the cocultures did not exhibithe morphology of mature neurons or glia, nor did theyxpress MAP2, a marker of mature neurons. This maye due to the short duration of incubation (maximum ofweeks) and a slower maturation rate for human-

erived cells. Other markers of neuronal development,uch as neurofilaments or b-III-tubulin (7), were not

tested in this series of experiments. The expression ofNeuN, but not MAP2 in BMSC-derived cells, is consis-tent with current knowledge regarding the timepointsof expression of neuronal proteins. MAP2 is expressedat a later developmental stage than NeuN. Immuno-histochemically detectable NeuN protein has been re-ported to first appear at developmental timepoints thatcorrespond with the withdrawal of the neuron from thecell cycle and/or with the initiation of terminal differ-entiation of the neuron (18). In contrast, the microtu-bule-associated proteins (MAP1, -2, and -3) undergo anumber of significant changes during development,with the expression of MAP2 considered to occur “late,”particularly between 10 and 20 days in the postnatalrat pup (22).

These results, along with those published by oth-ers, suggest that adult BMSC have a potentiallylarger developmental repertoire than previously ap-preciated. It should not be surprising that BMSC cangive rise to neural cells since the marker for neuralprecursors, nestin, was expressed in BMSC even inthe absence of the differentiation factors retinoicacid and BDNF. Moreover, BMSC have been shownto have the capacity to develop into hepatocytes,muscle, glia, and neuron-like cells under experimen-tal conditions (1, 5, 15, 19). Interestingly, these cellshave the capacity to migrate extensively and to ex-hibit a site-dependent differentiation. For example,the infusion via tail vein of retrovirus-labeled malebone marrow cells into irradiated female mice re-sulted in the influx of hematopoietic cells into brainover days to weeks (6). The marrow-derived cellswere found in cerebral cortex, hippocampus, thala-mus, brain stem, and cerebellum. Some marrow-de-rived cells expressed the microglial antigenic markerand others expressed the astroglial marker GFAP(6). BMSC injected into the lateral ventricles of new-born mice migrated extensively and differentiatedinto glial cells and occasionally into cells that ex-pressed neurofilament (13). Differentiation in theopposite direction has also been recently reported.Adult neural stem cells transplanted into irradiatedhosts developed into hematopoietic cells and a vari-

1

255BONE MARROW CELLS DIFFERENTIATE INTO NEURAL CELLS

ety of blood cells (2). Furthermore, cytokines nor-mally involved in hematopoeisis, such as IL-1, IL-11,and LIF have been shown to induce differentiation ofneural precursor cells into dopaminergic neurons(14).

A number of questions are raised by the presentfindings that should be addressed not with the aim toanswer them now, but rather with the aim to guideongoing and future research on these issues. AlthoughBMSC express nestin, a commonly used marker ofneural precursors, other cells may also transiently ex-press these intermediate filaments. Nestin is found inmyogenic cells, in newly formed endothelial cells ofextra- and intraembryonic blood vessels, epithelialcells of the developing lens, and hepatic stellate cells(16, 19). This, of course, raises the issue as to theembryonic origin of nestin-expressing cells outside ofthe nervous system, but more to the point, nestin is notan exclusive marker of neural precursors. Even theexpression by BMSC of a more mature neuronal pro-tein, neuron-specific nuclear protein, is insufficientproof that BMSC become neurons. The expression ofone or even two neuronal proteins does not prove thatthe cell bearing these “neuronal markers” is capable ofall the complex functions of a neuron. It will be impor-tant to determine whether longer incubation times re-sult in more mature neuron-like cells and whetherthese cells possess functional and electrophysiologicalcharacteristics of neurons. Additional fundamentalquestions remain to be answered. Identification andcharacterization of the stem cell or subfraction ofBMSC that gives rise to the neural precursor is not yetknown, although it is unlikely that hematopoietic stemcells are the source of the neural precursors. TheBMSC we utilized were depleted of hematopoietic stemcells and yet gave rise to cells with several neuralmarkers. Given the very small proportion of BMSCthat were induced to express neuronal or glial mark-ers, improved procedures need to be developed to en-rich the neural precursor population. The present find-ings, in the context of the rapidly expanding field ofstem cell biology, point to bone marrow as an invalu-able resource. Understanding the molecular mecha-nisms responsible for neuronal differentiation of thesecells will ultimately yield a readily available source ofneural cells for cellular therapies ranging from genetherapeutics to neural reconstruction in neurodegen-erative diseases, stroke, and trauma.

ACKNOWLEDGMENTS

This work was supported in part by the National Parkinson Foun-dation (Miami, FL), VA Merit Review Grant, Layton BioScience, Inc.(Sunnyvale, CA) and the Helen E. Ellis Research Endowment, Uni-versity of South Florida (Tampa, FL).

REFERENCES

1. Azizi, S. A., Stokes, D., Augelli, B. J., DiGirolamo, C., and Prockop,D. J. 1998. Engraftment and migration of human bone marrowstromal cells implanted in the brains of albino rats-similarities toastrocyte grafts. Proc. Nat. Acad. Sci. USA 95: 3908–3913.

2. Bjornson, C. R. R., Rietze, R. L., Reynolds, B. A., Magli, M. C.,and Vescovi, A. L. 1999. Turning brain into blood: A hemato-poietic fate adopted by adult neural stem cells in vivo. Science283: 534–537.

3. Chomczynski, P., and Sacchi, N. 1987. Single-step method ofRNA isolation by acid guanidinium thiocyanate-phenol-chloro-form extraction. Anal. Biochem. 162: 156–159.

4. Deacon, T., Dinsmore, J., Costantini, L. C., Ratliff, J., andIsacson, O. 1998. Blastula-stage stem cells can differentiateinto dopaminergic and serotonergic neurons after transplanta-tion. Exp. Neurol. 28–41.

5. Dinsmore, J., Ratliff, J., Deacon, T., Pakzaban, P., Jacoby, D.,Galpern, W., and Isacson, O. 1996. Embryonic stem cells differ-entiated in vitro as a novel source of cells for transplantation.Cell Transplant. 5: 131–143.

6. Eglitis, M. A., and Mezey, E. 1997. Hematopoietic cells differ-entiate into both microglia and macroglia in the brains of adultmice. Proc. Natl. Acad. Sci. USA 94: 4080–4085.

7. Fanarraga, M. L., Avila, J., and Zabala, J. C. 1999. Expressionof unphosphorylated class III beta-tubulin isotype in neuroep-ithelial cells demonstrates neuroblast commitment and differ-entiation. Eur. J. Neurosci. 11: 517–527.

8. Ferrari, G., Cusella-DeAngelis, G., Coletta, M., Paolucci E.,Stornaiuolo A., Cossu, G., and Mavilio, F. 1998. Muscle regen-eration by bone marrow-derived myogenic precursors. Science279: 1528–1530.

9. Fujita-Yamaguchi, Y., and Kathuria, S. 1988. Characterizationof receptor tyrosine-specific protein kinases by the use of inhib-itors. Staurosporine is a 100-times more potent inhibitor ofinsulin receptor than IGF-I receptor. Biochem. Biophys. Res.Commun. 157: 955–962.

0. Hatton, J. D., and Lin, L. 1992. Demonstration of specific neu-ronal cell groups in rat brain by b-galactosidase enzyme histo-chemistry. J. Neurosci. Methods 45: 147–153.

11. Kaufman, L. M., and Barrett, J. N. 1983. Serum factor support-ing long-term survival of rat central neurons in culture. Science220: 1394–1396.

12. Koblar, S. A., Turnley, A. M., Classon, B. J., Reid, K. L., Ware,C. B., Cheema, S. S., Murphy, M., and Bartlett, P. F. 1998.Neural precursor differentiation into astrocytes requires sig-naling through the leukemia inhibitory factor receptor. Proc.Natl. Acad. Sci. USA 95: 3178–3181.

13. Kopen, G. C., Prockop, D. J., and Phinney, D. G. 1999. Marrowstromal cells migrate throughout forebrain and cerebellum, andthey differentiate into astrocytes after injection into neonatalmouse brains [In Process Citation]. Proc. Natl. Acad. Sci. USA96: 10711–10716.

14. Ling, Z. D., Potter, E. D., Lipton, J. W., and Carvey, P. M. 1998.Differentiation of mesencephalic progenitor cells into dopami-nergic neurons by cytokines. Exp. Neurol. 149: 411–423.

15. Michel, P. P., Dandapani, B. K., Sanchez-Ramos, J., Efange, S.,Pressman, B. C., and Hefti, F. 1989. Toxic effects of potentialenvironmental neurotoxins related to 1-methyl-4-phenylpyri-dinium on cultured rat dopaminergic neurons. J. Pharmacol.Exp. Therapeutics 248: 842–850.

16. Mokry, J., and Nemecek, S. 1998. Immunohistochemical detec-tion of intermediate filament nestin. Acta Medica (HradecKralove) 41: 73–80.

17. Morse, W. R. 1938. Chinese Medicine. Hoeber, New York.

2

2

2

2

256 SANCHEZ-RAMOS ET AL.

18. Mullen, R. J., Buck, C. R., and Smith, A. M. 1992. NeuN, aneuronal specific nuclear protein in vertebrates. Development116: 201–211.

19. Niki, T., Pekny, M., Hellemans, K., Bleser, P. D., Berg, K. V.,Vaeyens, F., Quartier, E., Schuit, F., and Geerts, A. 1999. Class VIintermediate filament protein nestin is induced during activationof rat hepatic stellate cells [see Comments]. Hepatology 29: 520–527.

0. Petersen, B. E., Bowen, W. C., Patrene, K. D., Mars, W. M.,Sullivan, A. K., Murase, N., Boggs, S. S., Greenberger, J. S., andGoff, J. P. 1999. Bone marrow as a source of hepatic oval cells.Science 284: 1168–1170.

1. Prockop, D. J. 1997. Marrow stromal cells as stem cells fornon-hematopoietic tissues. Science 276: 71–74.

2. Riederer, B., and Matus, A. 1985. Differential expression ofdistinct microtubulbe-associated proteins during brain develop-ment. Proc. Nat. Acad. Sci. 82: 6006–6009.

3. Sanchez-Ramos, J. R., Michel, P., Weiner, W. J., and Hefti, F.1988. Selective destruction of cultured dopaminergic neuronsfrom embryonic rat mesencephalon: Cytochemical and mor-phological evidence. J. Neurochem. 50: 1934 –1944.

24. Vescovi, A. L., Parati, E. A., Gritti, A., Poulin, P., Ferrario,M., Wanke, E., Frolichsthal-Schoeller, P., Cova, L., Arcel-lana-Panlilio, M., Colombo, A., and Galli, R. 1999. Isolationand cloning of multipotential stem cells from the embryonichuman CNS and establishment of transplantable humanneural stem cell lines by epigenetic stimulation. Exp. Neurol.156: 71– 83.

25. Williams, L. R., Hilton, D. J., Pease, S., Willson, T. A.,Stewart, C. L., Gearing, D. P., Wagner, E. F., Metcalf, D.,Nicola, N. A., and Gough, N. M. 1988. Myeloid leukemiainhibitory factor maintains the developmental potential ofembryonic stem cells. Nature 336: 684 – 687.