Sonic hedgehog and retinoic acid synergistically promote sensory fate specification from bone marrow-derived pluripotent stem cells

Abstract

Recent studies demonstrated that stromal cells isolated from adult bone marrow have the competence of differentiating into neuronal cells in vitro and in vivo. However, the capacity of marrow stromal cells or mesenchymal stem cells (MSCs) to differentiate into diverse neuronal cell populations and the identity of molecular factors that confer marrow stromal cells with the competence of a neuronal subtype have yet to be elucidated. Here, we show that Sonic hedgehog (Shh) and retinoic acid (RA), signaling molecules secreted from tissues in the vicinity of peripheral sensory ganglia during embryogenesis, exert synergistic effects on neural-competent MSCs to express a comprehensive set of glutamatergic sensory neuron markers. Application of Shh or RA alone had little or no effect on the expression of these neuronal subtype markers. In addition, incubation of MSCs with embryonic hindbrain/somite/otocyst conditioned medium or prenatal cochlea explants promoted up-regulation of additional sensory neuron markers and process outgrowth. These results identify Shh and RA as sensory competence factors for adult pluripotent cells and establish the importance of interactions between adult pluripotent cells and the host microenvironment in neuronal subtype specification.

Resident in adult bone marrow is a cell population referred to as marrow stromal cells or mesenchymal stem cells (MSCs). MSCs possess several unique properties that make them a particularly attractive source of cells that could be used for cell-replacement therapy. Adult MSCs exhibit marked self-renewal capacity and the ability to differentiate into multiple cell types in vivo, establishing their pluripotent stem-like nature (1-3). By far the most prominent advantage of using MSCs over other cell types in cell-replacement therapy is their autologous characteristic. In addition, MSCs can easily be collected from patients under local anesthesia and propagate vigorously until they are committed to differentiate. Moreover, bone marrow transplantation has been proven to be a safe procedure in clinical venues for decades without any incidents of tumorigenesis in host tissues after transplantation. Despite these clear advantages, the full potential of MSC differentiation is not well understood. MSCs have been shown to differentiate into neuronal and glial cells both in vitro and in vivo. These MSC-derived cells expressing neuron-specific proteins display two types of voltage-gated ion channels, fast sodium and delayed rectifier potassium channels, characteristic of fully differentiated neurons (4-6). Thus far, dopaminergic, serotonergic, and GABAergic phenotypes have been observed in MSC-derived neurons (4, 5). However, no experimental protocol has been established for the efficient generation of a glutamatergic sensory neuron phenotype. In this study, we tested whether one or more signaling proteins secreted from the microenvironment surrounding sensory ganglia during embryogenesis has the ability to instruct MSCs to acquire morphological, biochemical, and molecular characteristics of peripheral nervous system (PNS) sensory neurons.

Materials and Methods

MSC Culture. Bone marrow was isolated from the femurs and tibias of 5- to 7-week-old C57BL/6 wild-type or TgN(ACTbEGFP) mice (The Jackson Laboratory). Epiphyses of femurs and tibias were removed, and the marrow was flushed out into a 100-mm culture dish. The isolated bone marrow, composed of hematopoietic and stromal cells, was maintained in high-glucose DMEM (Invitrogen) supplemented with 10% FBS (Invitrogen)/1% Gluta-Max (Invitrogen)/100 units/ml penicillin (Sigma) at 37°C with 5% CO₂ and passaged three to five times before being used for experiments. Hematopoietic and nonadherent cells were removed by changes of medium. For neural induction, cultured MSCs were first enzymatically detached from culture dishes and suspended in DMEM/10% FBS. The cells were plated on poly(D)-lysine-coated plastic culture dishes or eight-well chamber slides at 5 × 10⁴ cells per cm² and incubated for 2 days. At preconfluence, culture medium was replaced with preinduction medium composed of DMEM, 20% FBS, 10 ng/ml FGF2 (PeproTech, Rocky Hill, NJ), 2% B27 (Invitrogen), 50 μM Forskolin (Sigma), and 250 μM 3-isobutyl-1-methylxanthine (IBMX, Sigma) with or without 100 μM 2-mercaptoethanol (2-ME) and incubation continued for an additional 24 h. To initiate neuronal differentiation, the preinduction medium was removed, washed with PBS, and replaced with neuronal induction medium containing DMEM, 10 ng/ml FGF2, 1% insulin-transterrin-selenium supplements (Invitrogen), 2% B27, 5 μM Forskolin, 125 μM IBMX, 50 ng/ml BDNF (PeproTech) with or without 100 μM 2-ME, and one of the following reagents: (i) 400 ng/ml Sonic hedgehog (Shh) (R & D Systems), (ii) 0.5 μM all-trans retinoic acid (RA) (Sigma), (iii) Shh plus RA, (iv) embryonic day (E)10 mouse hindbrain/somite/otocyst conditioned medium, or (v) no factor (control). The cells were incubated for an additional 7 days. The presence or absence of 2-ME in the culture medium had no observable effect on the results, and results from both conditions were analyzed together.

3D Coculture. Organ of Corti tissues were removed from E18 C57BL/6 mice and placed in eight-chamber slide wells. For each culture, one organ of Corti was embedded into a gel mixture composed of rat type I collagen (BD Biosciences), 10× DMEM, and 10× neutralization solution (NaOH/NaHCO₃/Hepes) in a ratio of 8:1:1. After gelation at 37°C for 1 h, the gels were submerged into 200 μl of defined medium. Four to 5 hours after the start of incubation, GFP-positive MSCs at 1 × 10³ cells/5 μl were placed onto collagen by using a micropipette. The next day, the medium was replaced with neural induction medium and incubation continued for an additional 3 days.

Flow Cytometry. Polyclonal antibodies against Patched1, Smoothened, and RARα (Santa Cruz Biotechnology) were conjugated with Alexa Fluor 647 by using an antibody-labeling kit (Invitrogen). Purified MSCs were fixed, permeabilized with 0.1% Triton X-100, and incubated with one of the Alexa Fluor 647-conjugated antibodies overnight. Cell sorting and analysis were performed with a FACSCalibur flow cytometry system (BD Biosciences).

Conditioned Medium. The cervical region containing hindbrain, somite, and otocyst was removed from E10 wild-type mouse embryos. Six hindbrain/somite/otocyst tissues were placed in a well of a 12-well culture dish that contained 1 ml of serum-free defined medium. The tissues were incubated for 2 days at 37°C, and the culture medium was collected and filtered through a low protein-binding poly(vinylidene difluoride) membrane filter. The conditioned medium was diluted 1:1 in defined culture medium.

RT-PCR. The expression of panneuronal and sensory neuron marker genes was assayed by conventional or quantitative RT-PCR by a previously described procedure with modifications (7). At the end of cell culture, total RNA was isolated from MSCs by using the RNeasy Minikit (Qiagen, Valencia, CA) and then treated with TURBO DNase (Ambion, Austin, TX) to decrease the likelihood of DNA contamination. Single-stranded cDNA was synthesized by using Omniscript reverse transcriptase (Qiagen) and Oligo-dT primers. For nonquantitative RT-PCR, the resultant cDNA was amplified with Platinum taq DNA polymerase (Invitrogen) for 30 cycles. For quantitative RT-PCR, the cDNA samples were amplified on an ABI PRISM 7900HT Sequence Detection System (Applied Biosystems) by using the SYBR Green PCR Master Mix (Applied Biosystems). For each PCR reaction, a mixture containing cDNA template (5 ng), Master Mix, and forward and reverse primers (400 nM each) was first treated with uracil N-glycosylase at 50°C for 2 min and then underwent the following program: 1×, 95°C, 10 min; 45×, 95°C, 15 sec, 60°C, 1 min; 1×, 95°C, 15 sec, 60°C, 15 sec, 95°C, 15 sec (for melting curve analysis); 72°C, hold. After amplification, the PCR products were analyzed with the abi prism sequence detection software (Applied Biosystems). To check whether amplification yields PCR products with a single molecular weight, the PCR products were electrophoresed on 2% agarose gels containing ethidium bromide. In addition, melting curve analysis was performed to confirm the authenticity of the PCR products. To check for DNA contamination, PCR was run with cDNA samples by using an L27 (ribosomal housekeeping gene) primer pair, whose PCR product crosses an intron. To check the linearity of the detection system, a cDNA dilution series (1, 1/10, 1/100, and 1/1,000) was amplified with gene-specific primer pairs, and a correlation coefficient was calculated from the standard curve displaying threshold cycles (C_t) as a function of log₁₀ cDNA concentrations. The mRNA level for each probe (x) relative to L27 mRNA (internal control) was calculated as follows: mRNA(x%) = 2^{Ct (L27)-Ct(x)} × 100. Primer sequences are listed in Tables 1 and 2, which are published as supporting information on the PNAS web site.

Cell Proliferation Assay. Three days after plating cells on poly(D)-lysine-coated chamber slides, BrdUrd (BD Biosciences) at 10 μM was added to the culture medium and incubation continued for an additional 24 h. At the end of the incubation, the cultures were fixed with 4% paraformaldehyde for 30 min. The cells were denatured with 2 N HCl for 30 min. After blocking of nonspecific binding, the specimens were incubated with anti-BrdUrd antibody (BD Biosciences). BrdUrd immunoreactivity was visualized with an Alexa Fluor 488-conjugated anti-mouse IgG1. The specimens were counterstained with DAPI.

Immunohistochemistry. At the end of incubation, the cultures were fixed with 4% paraformaldehyde for 30 min. Chamber slides were incubated in blocking solution and then with a primary polyclonal and a monoclonal antibody together. Primary antibodies used in this study include: anti-Tau (Sigma), anti-Islet-1 (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City), anti-TUJ1 (Covance, Berkeley, CA), anti-NeuN (Chemicon), anti-MAP2 (Chemicon), 160 kDa antineurofilament (Chemicon), anti-GATA3 (Santa Cruz Biotechnology), anti-calretinin (Swant, Bellinzona, Switzerland), anti-GluR4 (Chemicon), and anti-Brn3a (Chemicon). Immunoreactivity with polyclonal and monoclonal antibodies was visualized by using an Alexa Fluor 488- (or 647) conjugated anti-rabbit IgG and Alexa Fluor 568- (or 647) conjugated anti-mouse IgG, respectively. For visualizing cellular nuclei, the specimens were counterstained with DAPI. Alexa Fluor 568 phalloidin (Molecular Probes) was used to visualize sensory hair cells in the organ of Corti. Some of the specimens were incubated with a primary antibody, and immunoreactivity was visualized by using the Elite ABC kit (Vector Laboratories) and Vector VIP or Vector SG as a substrate.

Results

Bone MSCs Constitutively Express Neural Stem Cell Markers and Acquire Neuronal Phenotypes After FGF2/Forskolin Stimulation. To test whether MSCs exhibit characteristics shared with neural vs. hematopoietic stem cells, we first examined the expression of nestin, Sox2, EgfR, FgfR1, and Notch1. These five genes are highly expressed in neural stem cells, but their expression levels are much lower in embryonic or hematopoietic stem cells (8). All of the neural stem cell markers, with the exception of Sox2, were detected in MSCs (Fig. 1I). To generate a neural progenitor pool from MSCs, we applied a mixture of FGF2, Forskolin, and 3-isobutyl-1-methylxanthine, reagents known to promote neural differentiation from neural stem cells or MSCs (5, 6, 9, 10). Within 72 h of application of the neural induction reagents, >70-80% of MSCs underwent dramatic changes in cell shape, from a flat mesenchymal cell shape to a round cell body with neurite-like processes (Fig. 1 A and B). To evaluate the rate of cell proliferation, BrdUrd incorporation assays were performed with MSCs before and after neural induction. Approximately 76% of unstimulated MSCs exhibited BrdUrd-positive cellular nuclei (Fig. 1 C₁, C₂, and J), whereas only 8% of MSCs that had been in neural-induction medium were positive for BrdUrd staining (Fig. 1 D₁, D₂, and J). These results suggest that unstimulated MSCs propagate vigorously, but that MSCs become postmitotic once they are committed to a neural lineage. Although the possibility of rapid cell death in stimulated MSCs cannot be ruled out, this is highly unlikely, given that there was little sign of apoptotic bodies in stimulated MSCs. To verify that these neuron-like cells are indeed neuronal, we performed immunohistochemical analysis for several early neuronal markers. None of the unstimulated MSCs showed positive staining for Islet1 or Tau (Fig. 1 E, G, and J), whereas the vast majority of stimulated MSCs were positive for Islet1 and Tau (Fig. 1 F, H, and J). These results indicate that in response to the neural induction reagents, ≈80% of MSCs acquire cellular characteristics of postmitotic neurons.

Fig. 1.

MSCs express neural stem cell markers and differentiate into postmitotic neurons in response to neural induction signals. (A-H) Immunohistochemical characterization of MSCs before (A, C₁, C₂, E, and G) and after (B, D₁, D₂, F, and H) neural induction. Shown are phase contrast (A and B), BrdUrd immunofluorescence (C₁ and D₁), DAPI (C₂ and D₂), Tau immunostaining (E and F), and Islet1 immunostaining (G and H). Note that cells with flat mesenchymal cell shape do not show positive staining for Tau or Islet1 (E and G). (Bars, 50 μm.) (I) RT-PCR analysis for neural stem cell markers in unstimulated MSCs. (J) Quantitative analysis of the percentage of BrdUrd-, Islet1-, or Tau-positive cells before (CT) and after (NI) neural induction. (K) RT-PCR analysis for panneuronal marker genes in MSCs before and after neural induction. CT, unstimulated MSCs; NI, MSCs after neural induction; PC, positive control, E10 otocyst or E14 brain cDNA (for nse, SCG10, and Tau); NC, no cDNA. (L) Quantitative RT-PCR for nse, SCG10, Tau, and BMP4 levels before (CT) and after (NI) neural induction.

To assess changes in gene expression before and after neural induction, RT-PCR analysis was performed to amplify expressed transcripts for panneuronal and sensory neuron markers. All of the panneuronal markers examined were undetectable or at very low levels in MSCs before neural induction (Fig. 1K and data not shown). After neural induction, however, nse, SCG10, and Tau levels in MSCs increased 10-, 8-, and 5-fold, respectively (Fig. 1 K and L). In contrast, the vast majority of sensory neuron markers remained undetectable after neural induction (Fig. 7, which is published as supporting information on the PNAS web site). Surprisingly, BMP4 was expressed at a high level in unstimulated MSCs but was down-regulated after neural induction (Fig. 1 K and L). In summary, the neural induction stimulation we used was able to initiate neural differentiation from MSCs but was not sufficient to turn on the expression of neuronal subtype markers.

Shh and RA in Combination, but Not Alone, Induce Expression of Glutamatergic Sensory Neuron Markers in MSCs. To direct MSCs expressing neuron-specific antigens to a more specific neuronal subtype, we focused on Shh and RA, because these signaling proteins have been shown to play important roles in genesis and differentiation of PNS sensory neurons (11-13). Our RT-PCR analysis showed that a receptor (Patched1) and downstream effectors (Gli2, Gli3, and Smoothened) for Shh, as well as RA nuclear receptors (RARα, -β, and -γ), are constitutively expressed in unstimulated MSCs (Fig. 2A). In addition, our flow cytometry analysis revealed that >97%, 98%, and 93% of MSCs express Patched1, Smoothened, and RARα proteins, respectively (Fig. 2B). Based on these results, we reasoned that MSCs may be sensitive to Shh and/or RA signaling. MSCs that had been incubated in preinduction medium for 1 day were transferred into neural-induction medium containing either Shh alone, RA alone, or Shh + RA and maintained for an additional 7 days. Thereafter, the cells were harvested, and mRNA expression was evaluated by RT-PCR (Fig. 3). None of the transcription factors examined was detectable in MSCs cultured with neural-induction medium or medium supplemented with Shh or RA alone. In contrast, substantial increases in Irx2 (4,000-fold) and Sox10 (400-fold), both of which are expressed in vestibulocochlear ganglia (VCG), trigeminal ganglia (TG), and dorsal root ganglia during development (14-17), were detected in MSCs grown for 7 days in the presence of Shh and RA together. In addition, GATA3, a zinc finger transcription factor expressed in VCG and TG (18-20), and P2X3, a sensory neuron-specific ionotropic purine receptor (21), were up-regulated 300- and 200-fold, respectively, in the presence of Shh and RA. However, Ret, a glial cell-line-derived neurotrophic factor coreceptor expressed in all sensory ganglia except VCG, was down-regulated during neuronal differentiation. As expected, RARβ was up-regulated by RA alone but not by Shh. These neural progenitors survived in vitro for >3 weeks when maintained in medium containing BDNF.

Fig. 2.

Receptors and downstream effectors for Shh and RA are constitutively expressed in MSCs. (A) RT-PCR analysis for Patched1 (Ptc1), Gli2, Gli3, Smoothened (Smo), and RARα, -β, and -γ mRNAs in unstimulated MSCs. (B) Flow cytometric scatter plots for Ptc1, Smo, and RARα in MSCs isolated from TgN(ACTbEGFP) mice. The majority of MSCs constitutively express Ptc1, Smo, and RARα and are GFP-positive (upper right of FACS plots). The presence of all these receptors and downstream targets strongly suggests that MSCs are responsive to Shh and/or RA signaling.

Fig. 3.

Sensory neuron marker genes are up-regulated in MSCs in the presence of Shh and RA. Quantitative RT-PCR analysis for transcription factors and membrane receptors in MSCs grown in basic culture medium (CT), neural induction medium (NI), neural induction medium containing Shh or RA alone (Shh, RA), or neural induction medium containing Shh and RA (S+R). Shh and RA in combination (S+R), but not alone, are able to up-regulate expression of GATA3, Sox10, Irx2, GluR3, GluR4, P2X3, and VGLUT1 in MSCs. Expression of calretinin and RARβ is regulated by RA but not by Shh. In contrast, Ret is down-regulated by NI.

Immunofluorescence analysis was performed to characterize changes in protein expression after neural induction (Fig. 4). None of the neuron-specific proteins examined was detected in unstimulated MSCs. In striking contrast, MSCs grown in neural induction medium containing Shh and RA displayed a range of mature neuronal antigens, including MAP2, TUJ1, NeuN, calretinin, and NF. GATA3 was also detected in the neural-competent MSCs.

Fig. 4.

Mature neuronal markers are expressed in MSCs in the presence of Shh and RA. Immunohistochemical characterization of MSCs grown in basic culture medium (CT, A₁-A_M) and in neural induction medium containing Shh and RA (S+R, B₁-D_M). TUJ1 (B₂), GATA3 (C₂), and NeuN (D₂) immunofluorescence is shown in red. Cells were also stained for a neuronal marker, MAP2 (B₁); calretinin (C₁); or NF (D₁) and the cellular nucleus marker DAPI (B₃, C₃, and D₃). Note that the cell density is much higher in the CT group because of active cell proliferation (A₃). Most cells in the S+R group are postmitotic. (Bar, 100 μm.)

To evaluate a neurotransmitter phenotype, we tested whether neural-competent MSCs express vesicular glutamate transporters (VGLUT)1-3. Of the three glutamate transporters, VGLUT1 and -2 are confined to glutamatergic neurons, whereas VGLUT3 is expressed in cells considered to release GABA (22). Expression of VGLUT1 was up-regulated 70-fold in MSCs grown with Shh and RA (Fig. 3), and a lesser but significant increase was observed for VGLUT2 (Fig. 8, which is published as supporting information on the PNAS web site). We also tested expression of three glutamate receptors (GluR2-4), because these AMPA receptors are expressed at high levels in type I spiral ganglia neurons as well as dorsal root ganglia neurons (23, 24). Our RT-PCR analysis showed that GluR2 was not detectable in MSCs regardless of their differentiation status (data not shown). In contrast, GluR3 and GluR4 were detected in differentiated MSCs that had been treated with Shh and RA (Fig. 3). GluR4 was up-regulated >100-fold in response to Shh and RA, whereas GluR3 that was constitutively expressed in untreated MSCs was also up-regulated by Shh and RA. In contrast, none of the GABA receptors examined, including GABA-Aα1, GABA-Aγ2, and GABA-Bβ1, was detected in MSCs (data not shown). The presence of VGLUT1/VGLUT2/GluR3/GluR4 and the absence of GABA receptors strongly suggest that MSCs have the competence to differentiate into glutamatergic but not GABAergic neurons. To test whether differentiated MSCs are sensitive to Ca²⁺ signals, we evaluated the expression of the calcium-binding protein calretinin. Undifferentiated MSCs expressed barely detectable levels of calretinin (Fig. 3). After neuronal differentiation with Shh and RA, however, the calretinin mRNA level increased dramatically. In addition, calretinin protein became detectable in MSCs after the application of Shh and RA (Fig. 4C). The up-regulation of the calcium-binding protein, along with the predominant expression of AMPA receptors, suggests that differentiated MSCs exhibit increased intracellular Ca²⁺ levels in response to glutamate.

Involvement of Additional Signaling Proteins in Tissue-Specific Differentiation of MSCs. Although the majority of sensory neuron markers examined were up-regulated in MSCs that were treated by Shh and RA, we were unable to detect the POU-domain transcription factor Brn3a (25). To test the possibility that hitherto unknown proteins in the microenvironment surrounding sensory ganglia have effects on Brn3a expression, we incubated MSCs that had been exposed to neural induction medium with E10 hindbrain/somite/otocyst conditioned medium. Incubation with such conditioned medium for 3 days induced a 150-fold increase in Brn3a, which was accompanied by sharp up-regulation of neuroD and neurogenin1, bHLH genes essential for cranial sensory neurogenesis (Fig. 5). In contrast, expression levels of GATA3 and Sox10 were not changed by the conditioned medium. To investigate interactions between MSCs and embryonic sensory tissues, we cocultured MSCs with an E18 organ of Corti, the target tissue for developing cochlear ganglion neurons. In the presence of an organ of Corti explant, MSCs survived and propagated vigorously, but upon exposure to neural induction reagents, they withdrew from the cell cycle and acquired neuron-like cell morphology (Fig. 6). These cellular characteristics exhibited by MSCs in 3D cocultures were similar to those observed in 2D MSC primary cultures. In addition, the MSCs began expressing defined sensory neuron markers, including Brn3a, GluR4, and calretinin (Fig. 6 B, D, and F). Furthermore, MSCs extended processes toward sensory hair cells in the explants, suggesting positive interactions between MSCs and the organ of Corti (Fig. 6 C and E).

Fig. 5.

Hindbrain/somite/otocyst conditioned medium up-regulates Neurogenin1, NeuroD, and Brn3a, but not GATA3 or Sox10, in neural-competent MSCs. Quantitative RT-PCR analysis for Neurogenin1 (Ngn1), NeuroD, Brn3a, GATA3, and Sox10 in MSCs in basic culture medium (CT), neural induction medium (NI), neural induction medium containing Shh and RA (S+R), or E10 hindbrain/somite/otocyst conditioned medium (CM). (Lower Left) PCR products for Ngn1, NeuroD, and Brn3a from MSCs in CT, S+R, or CM. Positive (PC, E10 otocyst cDNA) and negative (NC, no cDNA) controls are also shown.

Fig. 6.

Sensory tissues promote neuronal differentiation and process outgrowth from MSCs. Low- (A and B) and high- (C-F) magnification photographs of 3D collagen culture with MSCs and an organ of Corti explant. (A, C, and E) GFP-positive MSCs (green) survive and extend their neurite-like processes (arrows in C and E) toward phalloidin-stained sensory hair cells (red) in the organ of Corti. Cellular nuclei were stained with DAPI (blue). (B, D, and F) These MSCs cocultured with an organ of Corti explant for 3 days express sensory neuron markers, including Brn3a (B), GluR4 (D), and calretinin (F). OHCs, outer hair cells; IHCs, inner hair cells. [Bars, 500 (B) and 50 (D and F) μm].

Discussion

In this study, we have presented evidence that Shh and RA exert synergistic effects on neural-competent MSCs to induce expression of a comprehensive set of genes and proteins that define PNS sensory neurons. Genes that were up-regulated >70-fold when compared with the expression level in untreated MSCs include Irx1, Irx2, Sox10, GATA3, GluR4, P2X3, VGLUT1, and calretinin, all of which are expressed in VCG, TG, or dorsal root ganglia during normal embryonic development. Shh or RA alone had little or no effect on the expression of these glutamatergic sensory neuron marker genes. Because both Shh and RA are released during embryonic development from the hindbrain and somites in the vicinity of sensory ganglia, our results strongly suggest that extrinsic signals, most likely diffusible signaling proteins, in the embryonic microenvironment surrounding sensory ganglia have the capacity to instructively promote the expression of the transcription factors and glutamatergic markers in adult pluripotent stem-like cells.

Although several transcription factors expressed in sensory neurons were up-regulated by Shh and RA, they were unable to induce Brn3a. However, incubation with hindbrain/somite/otocyst conditioned medium led to a 150-fold increase in Brn3a in MSCs. Furthermore, incubation with an embryonic inner ear explant resulted in the induction of the Brn3a protein in MSCs. These results strongly suggest that soluble protein(s) in the embryonic microenvironment, other than Shh and RA, have an instructive role in Brn3a expression. Although the identity of the signaling protein(s) that controls Brn3a expression in MSCs has yet to be determined, recent evidence suggests that Wnt/β-catenin signaling may be involved in this process. Wnt1, in the presence of β-catenin, was shown to instruct neural crest stem cells to differentiate into Brn3a-positive sensory neurons (26). Consistent with this, DKK1, a canonical Wnt signaling inhibitor, was able to suppress the hindbrain/somite/otocyst conditioned medium-induced up-regulation of Brn3a (T.K., unpublished observations). Moreover, Brn3a protein was induced in MSCs that had been cultured with embryonic inner ear sensory epithelium that synthesizes Wnt4, -6, and -7A (27-29). Interestingly, the up-regulation of Brn3a in MSCs was accompanied by up-regulation of neurogenin1 and neuroD but not of GATA3 or Sox10, both of which are up-regulated by Shh and RA. It is tempting to speculate that Shh/RA signaling and Wnt signaling regulate distinct and mutually exclusive sets of transcription factors. Accordingly, it seems plausible to hypothesize that Brn3a suppresses expression of GATA3 and Sox10 in MSCs, because Shh and RA, as well as Wnts, are likely to be present in the hindbrain/somite/otocyst conditioned medium. Support for this hypothesis comes from a recent study, in which targeted elimination in the Brn3a gene was shown to induce a substantial increase in GATA3 mRNA in the TG (20). In addition, during differentiation, sensory neural progenitor cells lose Sox10 immunoreactivity while acquiring Brn3a expression (26). Inclusion of recombinant Shh and RA in the hindbrain/somite/otocyst conditioned medium did not increase the GATA3 level in MSCs (data not shown), further supporting our hypothesis. It should be noted, however, that gene regulatory processes in adult pluripotent cells might be different from those in PNS neural progenitors. It also remains to be seen whether other types of pluripotent or stem cells are also sensitive and respond to Shh/RA signaling in the same manner as MSCs. Nevertheless, our present results have significant clinical implications in establishing an autologous cell transplantation therapy to cure PNS neuropathy. Pretreatment of MSCs with neural induction reagents in conjunction with Shh and RA could conceivably increase the number of MSCs that exhibit morphologic and phenotypic characteristics of glutamatergic sensory neurons. In particular, it seems feasible to generate different sensory phenotypes (VCG, TG, or dorsal root ganglia neurons) by changing the balance between Shh/RA and Wnts during neural induction. Alternatively, coapplication of Shh and RA into a recipient tissue could greatly enhance the ability of donor cells to differentiate into neurons expressing a defined set of sensory neuron markers. Validation of these hypotheses awaits further in vivo transplantation experiments. The present results also demonstrated that exposure of pluripotent cells to a combination of secreted proteins is sufficient to up-regulate the expression of proneural genes and transcription factors. This suggests that tissue-specific neurons can be generated without genetic engineering of pluripotent MSCs.

How does Shh and RA signaling converge and exert effects on transcriptional regulation of sensory neuron marker genes in MSCs? Although several lines of evidence suggest that Shh and RA signaling communicates during embryonic development (30, 31), the underlying mechanisms remain obscure. A recent study (32) provided evidence that Tlx3 and Tlx1, Tlx-class homeobox genes, promote glutamatergic differentiation in spinal cord neural progenitors. Tlx3 was shown to control the expression of several glutamatergic markers while suppressing the GABAergic marker Pax2. We tested whether Tlx3 or Tlx1 is involved in Shh- and RA-mediated glutamatergic differentiation from neural-competent MSCs, but neither of these genes was detected in our cell preparations (data not shown). Alternatively, the fusion of donor bone marrow cells to neuronal cells in the host tissues has been suggested to account for neuronal differentiation from bone marrow-derived stem cells (33, 34), whereas evidence against the cell fusion phenomenon has also been reported recently (35). The possibility of MSC fusion to neuronal cells is highly unlikely in this study, because the primary MSC cultures used did not contain any neuronal cells at the start of incubation. Further investigation is required to elucidate molecular events downstream of Shh and RA that regulate coordinated expression of sensory neuron marker genes and the emergence of the glutamatergic phenotype. Future studies should also include the validation of a role for Wnt signaling in Brn3a activation, which could result in suppression of GATA3 and Sox10 in MSCs.