Neuronal Differentiation of Bone Marrow-derived Stromal Stem Cells Involves Suppression of Discordant Phenotypes through Gene Silencing*

Tissue engineering involves the construction of transplantable tissues in which bone marrow aspirates may serve as an accessible source of autogenous multipotential mesenchymal stem cells. Increasing reports indicate that the lineage restriction of adult mesenchymal stem cells may be less established than previously believed, and stem cell-based therapeutics await the establishment of an efficient protocol capable of achieving a prescribed phenotype differentiation. We have investigated how adult mouse bone marrow-derived stromal cells (BMSCs) are guided to neurogenic and osteogenic phenotypes. Naïve BMSCs were found surprisingly active in expression of a wide range of mRNAs and proteins, including those normally reported in terminally differentiated neuronal cells and osteoblasts. The naïve BMSCs were found to exhibit voltage-dependent membrane currents similar to the neuronally guided BMSCs, although with smaller amplitudes. Once BMSCs were exposed to the osteogenic culture condition, the neuronal characteristics quickly disappeared. Our data suggest that the loss of discordant phenotypes during BMSC differentiation cannot be explained by the selection and elimination of unfit cells from the whole BMSC population. The percent ratio of live to dead BMSCs examined did not change during the first 8–10 days in either neurogenic or osteogenic differentiation media, and cell detachment was estimated at <1%. However, during this period, bone-associated extracellular matrix genes were selectively down-regulated in neuronally guided BMSCs. These data indicate that the suppression of discordant phenotypes of differentiating adult stem cells is achieved, at least in part, by silencing of superfluous gene clusters.

Tissue engineering involves the construction of transplantable tissues in which bone marrow aspirates may serve as an accessible source of autogenous multipotential mesenchymal stem cells. Increasing reports indicate that the lineage restriction of adult mesenchymal stem cells may be less established than previously believed, and stem cell-based therapeutics await the establishment of an efficient protocol capable of achieving a prescribed phenotype differentiation. We have investigated how adult mouse bone marrow-derived stromal cells (BMSCs) are guided to neurogenic and osteogenic phenotypes. Naïve BMSCs were found surprisingly active in expression of a wide range of mRNAs and proteins, including those normally reported in terminally differentiated neuronal cells and osteoblasts. The naïve BMSCs were found to exhibit voltage-dependent membrane currents similar to the neuronally guided BMSCs, although with smaller amplitudes. Once BMSCs were exposed to the osteogenic culture condition, the neuronal characteristics quickly disappeared. Our data suggest that the loss of discordant phenotypes during BMSC differentiation cannot be explained by the selection and elimination of unfit cells from the whole BMSC population. The percent ratio of live to dead BMSCs examined did not change during the first 8 -10 days in either neurogenic or osteogenic differentiation media, and cell detachment was estimated at <1%. However, during this period, bone-associated extracellular matrix genes were selectively down-regulated in neuronally guided BM-SCs. These data indicate that the suppression of discordant phenotypes of differentiating adult stem cells is achieved, at least in part, by silencing of superfluous gene clusters.
Bone marrow represents an abundant source of renewing stem cells with potential for developing into multiple lineages (1)(2)(3)(4)(5)(6). Bone marrow transplantations in humans and animals have led to an understanding that the donor's bone marrowderived stromal cells (BMSCs) 1 can successfully integrate into a wide range of highly specialized tissues of the recipient in vivo. Transplanted BMSCs have been shown to fuse with differentiated resident cells, such as skeletal and cardiac muscles (7)(8)(9)(10)(11)(12)(13), liver (14,15), and neuronal cells (16 -21). However, Houghton et al. (22) have recently shown that the transplanted BMSCs can develop gastric epithelial cells that contain a single nucleus, suggesting that the formation of heterokaryons is not necessarily a prerequisite for BMSC differentiation. It must be noted that their study has further demonstrated that transplanted BMSCs can develop epithelial cancers under chronic Helicobacter infection (22). Regenerative medicine strategy using undifferentiated BMSC transplantation may thus require stringent monitoring.
Bone marrow aspirate can serve as an accessible source of cellular components for tissue engineering strategies. Increasing reports on in vitro differentiation protocols for adult mesenchymal stem cells indicate that the lineage restriction may be less stringent than previously believed. Experiments with multipotent adult stem cells in vitro suggest that the microenvironment contributes substantially to terminal differentiation (12,14,23). For example, BMSCs show the potential to adopt widely different end points not only of their well characterized mesenchymal derivatives, such as osteoblasts, chondroblasts, and adipocytes (24 -26), but also of ectoderm-derived neural cells (27)(28)(29)(30)(31)(32)(33). The in vitro trans-lineage differentiation capability of stem cells should broaden the engineering application to a wide range of tissues, each of which awaits the establishment of an efficient protocol that stably guides them to a prescribed terminal differentiation.
Both rodent and human BMSCs can be rapidly induced to differentiate into neurons in a defined in vitro microenvironment (34). Shortly after the exposure to the neurogenic culture condition, BMSCs begin to develop characteristic neuron-like morphologies, such as processes resembling axons and dendrites (neurites). These cells also express genes and proteins that are normally associated with neuronal cells. We used the adult mouse BMSC model undergoing ectodermal/neurogenic or mesodermal/osteogenic differentiation to elucidate the molecular mechanism regulating the in vitro trans-differentiation of BMSCs. In this study, we found that naïve BMSCs in vitro, rather than expressing only immature precursor genes, indeed express a broad range of gene clusters characteristic of the advanced phenotypes of the diverse potential end points. Guided differentiation may thus employ a unique conversion from broad to specific phenotype expression through silencing of discordant gene clusters.

MATERIALS AND METHODS
BMSC Isolation and Cell Culture-Mouse BMSCs were obtained by a protocol approved by the UCLA Animal Research Subjects Committee. Bone marrow cells were collected from femurs of adult male C57BL/6 mice as described previously (31) and maintained in "control medium" consisting of minimum essential medium ␣ supplemented with 15% fetal bovine serum (Invitrogen), 100 units/ml penicillin, 100 g/ml streptomycin, and 250 ng/ml amphotericin B. The non-adherent cell population was removed after 48 h, and the adherent BMSC layer was washed twice with fresh medium. The cells were then continuously cultured for 1 week.
For neurogenic induction, subconfluent BMSCs were cultured in the control medium supplemented with 1 mM ␤-mercaptoethanol (Sigma) for 24 h followed by culture in neurobasal medium supplemented with B27 and 20 ng/ml of brain-derived neurotrophic factor (Invitrogen). For osteogenic induction, subconfluent BMSCs were cultured in the control medium supplemented with 10 Ϫ8 M of dexamethasone, 10 mM of ␤-glycerophosphate, and 50 g/ml of ascorbic acid. Day 0 refers to the day when the neurogenic or osteogenic induction commences.
RT-PCR Analysis-Total RNA was extracted with TRIzol reagent (Invitrogen) and the RNeasy Mini Kit (Qiagen, Hilden, Germany). After DNase I treatment (Ambion, Austin, TX), cDNAs were synthesized from 1 g of total RNA using Omniscript reverse transcriptase (Qiagen). The cDNA target was amplified by PCR using Ex TaqDNA polymerase (Takara Biotechnology, Shiga, Japan) following the manufacturer's recommendations. The PCR conditions and primer pairs used are given in Table I. For semiquantitative analysis, the intensity of bands was quantified under ultraviolet light (Eagle Eye II; Stratagene, La Jolla, CA) and normalized with respect to those for ␤-actin mRNA.
Electrophysiology: Patch Clamp Recordings-Whole-cell voltage clamp recordings were made with a patch clamp amplifier (Cairn Optopatch, Cairn Research LTD, Kent, UK) using jClamp software (www.scisoftco.com) and a digital data acquisition board (6052E, National Instruments, Austin, TX). Patch electrodes had a resistance of 3-4 M⍀, and series resistance was below 20 M⍀. The standard internal pipette solution contained (in mM) 10 NaCl, 130 KCl, 1 MgCl 2 , 10 HEPES, 0.5 EGTA, 4 Mg-ATP, and 0.6 Na-GTP and was adjusted to pH 7.2. The standard external solution contained (in mM) 119 NaCl, 5 KCl, 30 glucose, 25 HEPES, 2 CaCl 2 , and 2 MgCl 2 and was adjusted to pH 7.2. When indicated, tetrodotoxin (TTX, 100 nM) and tetraethylammonium chloride (TEA, 10 mM) were used to block sodium and potassium channels, respectively. Cells were depolarized for 300 ms from a holding potential of Ϫ80 mV to the indicated test potentials. Leak subtraction was performed using a standard P/4 protocol (i.e. the hyperpolarizing subtraction pulse was four times smaller than the depolarizing test pulse) and the average current at the end of the depolarization was measured.
Alkaline Phosphatase (ALP) Assay and von Kossa Staining-Cells were fixed in 10% buffered formalin phosphate (Fisher) for ALP assay and von Kossa staining. ALP-positive cells were stained by incubating with 1.8 mM fast red TR (Sigma) and 0.9 mM naphthol AS-MX phosphate (Sigma) in 120 mM Tris buffer (pH 8.4) for 30 min at 37°C. Nodule mineralization was visualized by treating with 5% silver nitrate solution (Sigma) under ultraviolet light for 30 min, followed by 5% sodium thiosulfate solution (Sigma) for 5 min. The image was scanned, and the area staining positive for ALP or mineralization was analyzed by Image-Pro Plus version 2.0 (Media Cybernetics, Silver Spring, MD).
Immunocytochemistry-After fixation, cells were incubated for 20 min in blocking solution (1% bovine serum albumin and 0.1% Triton X-100 in phosphate-buffered saline) and then washed and incubated overnight at 4°C with a primary antibody in phosphate-buffered saline containing 1% bovine serum albumin. The primary antibodies were rabbit polyclonal anti-Trk A (1:50; Santa Cruz Biotechnology, Santa Cruz, CA) and mouse anti-␤-tubulin III (TUJ1) (1:250; Covance, Berkeley, CA). To assess nonspecific staining, rabbit and mouse immunoglobulin G (IgG) (1:100; Santa Cruz Biotechnology) were used at the same concentration as the primary antibodies of Trk A and TUJ1, respectively. The cells were then washed and incubated for 1 h at 37°C with the secondary antibody, either fluorescein isothiocyanate-conjugated goat anti-rabbit IgG (1:100; Santa Cruz Biotechnology) or Alexa Fluor 488 goat anti-mouse IgG (1:500; Molecular Probes, Eugene, OR).
Cell Viability Assay-Living and dead cells were distinguished using fluorescence microscopy according to the procedure provided with the live/dead Viability/Cytotoxicity Kit (Molecular Probes, Eugene, OR). Two fields at 200ϫ magnification were randomly selected from each well (8-well chamber slide), and the numbers of labeled living and dead cells were counted. The percentage of live cells in each field was calculated, and mean values Ϯ S.D. of quadruplicate wells were obtained with for each condition.
To examine the detached cell population during the differentiation process, floating cells in day-9 osteoblastic differentiation medium were collected by centrifugation. The cells were washed with 1ϫ phosphatebuffered saline and stained for apoptosis using fluorescein isothiocyanate-annexin V/propidium iodide (Apoptosis Detection Kit, Coulter Immunotech, Fullerton, CA). After the cell density was determined using a hemacytometer, the cells were characterized by flow cytometry to determine the percentage of live and dead cells in the floating cell population.

RESULTS
Mouse BMSCs maintained in non-inducing medium (control medium) exhibited three distinct morphologies: large flat cells, oval-to spindle-shaped cells, and small round cells (Fig. 1A,  Day 0). Cultured in neurogenic induction medium, most of the spindle-shaped and large flat BMSCs retracted their extensions, and round cell bodies became predominant after 1 day (Fig. 1A, Day 1). Subsequently, the cells acquired a neuron-like cell shape with processes similar to axons and dendrites with growth cones, as described previously (28, 37-39) (Fig. 1A,  Days 2 and 10). The BMSCs in the neurogenic induction medium expressed mRNA species encoding ␤-tubulin III and nes- tin; however, little expression of Pax6, glial fibrillary acidic protein (GFAP), and myelin basic protein (MBP) was observed (Fig. 1B). The membrane currents of neurogenically induced BMSCs on day 4 displayed distinct voltage dependence with a peak inward current at a depolarized voltage of ϳϩ40 mV. TTX eliminated the inward current and unmasked an outward current. The combined application of TEA and TTX almost completely eliminated the voltage-dependent currents (Fig. 1C). In osteogenic induction medium, BMSCs exhibited predominantly large, flat cell shapes on day 8, whereas BMSCs in the control medium remained in the original cell shapes (Fig. 2A). The strong expression of osteocalcin, osteopontin, and ALP was observed by day 8 (Fig. 2, B and C), and mineralized nodules were formed covering 19.4% of the culture area by day 15 (Fig. 1B).
While establishing these neurogenic and osteogenic differentiation protocols, we have repeatedly observed that naïve BM-SCs cultured in a control medium expressed genes indicative of both neuronal (␤-tubulin III and nestin; Fig. 1B) and osteoblastic (osteocalcin, osteopontin, and ALP; Fig. 2B) phenotypes. In addition, the majority of these naïve, uninduced BMSCs also expressed the neuronal proteins Trk A and ␤-tubulin III. However, their spindle-shaped cell bodies did not display neuronal (neurites) or osteoblastic (mineralized nodules) characteristics (Fig. 3). Staining intensity for neuronal markers in naïve BM-SCs was equivalent to that of neurogenically induced BMSCs, which developed distinct, neurite-like processes (Fig. 3, C and  G), but no neuronal markers were detected in osteogenically induced BMSCs (Fig. 3, D and H). We further examined the electrophysiological properties of control, neurogenically, and osteogenically induced BMSCs using whole-cell voltage clamp recordings. We were unable to record from osteogenic cells in the later stages in culture probably because of substantial modification of the cellular membrane during early calcification. Therefore, we limited our recordings to cells differentiated for 6 days. Neuronal BMSCs expressed robust, voltage-activated inward currents (Figs. 1C and 3I). Uninduced BMSCs also expressed voltage-activated currents, although of smaller amplitude and with considerable variation between cells (Fig.  3I). When BMSCs were exposed to the osteogenic differentiation medium, no voltage-dependent currents were detectable at any of the voltages tested (Fig. 3I).
The cell vitality of the BMSCs was evaluated by measuring intracellular esterase activity and plasma membrane integrity (Fig. 4, A and B). The percent ratio of live and dead BMSC was unchanged until 8 -10 days of culture in neurogenic and osteogenic differentiation media, respectively, followed by a sharp decrease of the live cell ratio. BMSCs in control medium main-tained a constant live/dead cell ratio throughout the study period (Fig. 4A). At day 9 in osteogenic differentiation medium, it was found that Ͻ1% of the anchored cells detached, and 89.9% of the detached cells showed signs of apoptosis.
Uninduced BMSCs constitutively expressed nestin, ␤-tubulin III, and osteocalcin during the early culture period. The steady state expression levels of nestin and ␤-tubulin III were decreased by as much as 85 and 20%, respectively, when BMSCs were cultured in osteogenic differentiation medium. BMSCs in neurogenic medium, on the other hand, showed a 47% down-regulation in osteocalcin expression by 8 days (Fig. 4C, arrows).
Custom-made cDNA microarray chips containing 58 ECMand bone-related genes and 96 housekeeping genes (35,36) were used to further examine the extent of gene silencing during the early differentiation period. Total RNA samples of osteogenic or neurogenic BMSCs, taken at day 6, labeled with the fluorescent marker Cy3 were compared with the total RNA (labeled with Cy5) from non-induced BMSCs on the same culture day. Analysis of the Cy3/Cy5 fluorescence ratio indicated that, when BMSCs underwent osteogenic differentiation, ECM/ bone-related genes were up-regulated. In contrast, these genes were silenced when BMSCs underwent neurogenic differentiation (Fig. 4D). DISCUSSION Our experiments showed that the same batch of adult mouse BMSCs can develop ectodermal/neuronal and mesenchymal/ osteoblastic phenotypes. The differentiation process of various cellular phenotypes was monitored by observing cell morphology and the expression of mRNA, proteins, and functional cellular properties characteristic of the guided differentiated lineages. We demonstrated that neuronal cells derived from BMSCs expressed mRNA species encoding ␤-tubulin III (an early neuronal marker) and nestin (a neuronal stem cell marker) (28,31,36,38,39). However, little expression of the neuroectodermal marker Pax6 and the glial markers GFAP and MBP was observed (Fig. 1B). These BMSCs may differentiate directly to neuroprogenitor cells without typical neuroectodermal passages under this culture condition. The neurogenically induced BMSCs on day 4 displayed distinct voltagedependent membrane currents with a peak inward current at a depolarized voltage of ϳϩ40 mV. TTX, a specific blocker of voltage-gated sodium channels, eliminated the inward current and unmasked an outward current. The combined application of the potassium channel blocker TEA and TTX almost completely eliminated all voltage-dependent currents. This suggests that the neurogenically induced BMSCs express functional sodium and potassium channels. The expression of neuronal markers, such as nestin, TrkA, and ␤-tubulin III and the concurrent demonstration of voltage-gated currents are a hallmark of neurogenic differentiation (37,38,40,41) in embryonic and neural stem cells.
In the literature, the differentiation process of BMSCs has been followed by the monitoring expression of mRNA, proteins, and functional cellular properties characteristic of the guided differentiated lineages (25,29,42). These studies have followed the time-dependent expression profile of these differentiation markers above baseline levels. Our results suggest, however, that the expression of osteogenic and neuronal molecular markers as well as the voltage-dependent currents are a constitutive cellular property of uninduced, naïve BMSCs in the absence of differentiation-inducing stimuli (Fig. 3). A recent paper by Tondreau et al. (43) suggested that, after several passages, undifferentiated BMSCs expressed neuronal genes and proteins examined by RT-PCR and Western blot. Taken together, these findings indicate that nonspecific gene expression can be an intrinsic property of adult stem cells.
Isolated BMSCs have been observed in various morphologies representing the heterogeneous nature of BMSCs composed of mesenchymal precursors, including the recently described recycling/rapidly self-renewing stem cells (44,45) and hematopoietic precursors (46). It must be noted that different cell shapes have also been reported in human clonal BMSCs (44,45,47), although the broad and flattened shape eventually predominates after repeated cell divisions (48 -51). In the BMSC population, some stem cells may be true pluripotential cells, i.e. multipotential adult progenitor cells (39), whereas others may have already undergone lineage-specific differentiation at various stages. In our study, the BMSCs were not initially screened and contained a mixture of cells ranging from multipotential adult progenitor cells to mesenchymal stem cells with various differentiation commitments. The expression of a mixture of lineage-specific phenotypes by the BMSCs may well represent the collective phenotypic profiles of these heterogenic cells. It has been postulated that hematopoietic cell differentiation undergoes positive and negative selection among various cell populations (52)(53)(54). It seemed possible that the down-regulation of unrelated phenotype markers and cellular properties resulted from negative selection. If this were the case, the in vitro differentiation of BMSCs should, in part, employ a selection and elimination process in which unfit cells undergo apoptosis. We addressed this possibility by a simple cell live/dead assay (Fig. 4). To our surprise, cell vitality was well maintained until the second week in both osteogenic and neurogenic differentiation media. A sudden decrease of cell vitality occurred at days 8 and 10 in neurogenic and osteogenic culture condition, respectively. Therefore, negative selection may occur in the late differentiation stages. However, during the early differentiation stages, the negative selection of unfit cells alone cannot explain the down-regulation of unrelated cellular properties. Moreover, during the early differentiation stages, Ͻ1% of cells were found detached and floating in the differentiation medium. Most of these detached cells underwent apoptosis. However, it was not determined whether apoptotic cells exfoliated from the culture dish or the loss of integrin-mediated anchorage due to mechanical dislodgement induced apoptosis, described as anoikis (55). Nonetheless, it is unlikely that selective loss of subpopulations of BMSCs through detachment contributed to the loss of marker gene expression.
It is possible then, that the whole BMSC differentiation process entails a conversion from broad to phenotype-specific gene expression. We showed that such conversion may involve suppression of non-phenotype-specific genes (Fig. 4). During the embryogenesis, the ultimate totipotent cells gave rise to all cell types. Animal development can therefore be described as a progressive loss of totipotency, and then pluripotency, and ultimately differentiation into specific cell types. Using large scale microarray analysis, Tanaka et al. (56) suggest that totipotency of embryonic stem cells requires a set of genes not expressed in other cell types, whereas lineage-restricted stem cells, such as extra-embryonic-restricted trophoblast stem cells, express genes predictive of their differentiated lineage. Adult BMSCs are not totipotent stem cells but are considered lineage-restricted multipotent stem cells, although the phenotype restriction may not be as narrow as previously thought.
The mechanism that regulates adult stem cell differentia-tion, by pruning a repertoire of differentiation-unrelated genes down to lineage-specific characteristics, may share the underlying mechanism of positive and negative gene regulation during development (57). However, the rapid loss of the steady state level of gene expression activity observed in this study may present a unique property of adult stem cell undergoing phenotype differentiation in the in vitro environment. For tissue engineering applications, it is important to lead stem cells to the targeted end point, which may be guided not only by up-regulating tissue phenotype-sensitive genes but also by suppressing discordant gene expression. We speculate that this transcriptional down-regulation could involve suppressor-like nuclear factors, rapid DNA methylation, or RNA interference. In any event, the postulated "gene pruning" differentiation mechanism may provide the basis for stem cell-based biotechnology and for potential applications in tissue engineering.
Acknowledgments-We thank Mr. Devang Thakor of the UCLA Neuroengineering Program for his critical review of this manuscript and Dr. Linda Dubin, UCLA School of Dentistry, for editorial assistance. This investigation was conducted in a facility constructed with support from Research Facilities Improvement Program Grant Number C06 RR014529 from the National Center for Research Resources, National Institutes of Health.
FIG. 4. Down-regulation of the steady state expression levels of phenotype-unrelated genes during the early differentiation stages. A, percentage of live cells in control (Con), osteogenic (Os), and neurogenic (Neu) cultures. B, intracellular esterase activity and plasma membrane integrity differentiated the live (green fluorescent) and the dead (red fluorescent) BMSCs in osteogenic and neurogenic cultures. C, time course RT-PCR analysis for lineage marker transcripts in control, osteogenic, and neurogenic cultures. Arrows indicate reduction of gene expression levels compared with non-induced cells. D, cDNAs from osteogenic, neurogenic, and control cultures on day 6 were hybridized on a microarray spotted with various ECM-and bone-related genes (black dots). Ninety-six housekeeping genes that were co-spotted on the microarray were used as a standardization control.