Author contributions: V.G., Z.Z.P., and S. M.: conception and design, financial support, collection, interpretation and assembly of data, manuscript writing; B.B., N.D.S.-K., H.A., and G.N.F.: collection, interpretation and assembly of data.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLS EXPRESS August 26, 2010.
Production of neurons from non-neural cells has far-reaching clinical significance. We previously found that myoblasts can be converted to a physiologically active neuronal phenotype by transferring a single recombinant transcription factor, REST-VP16, which directly activates target genes of the transcriptional repressor, REST. However, the neuronal subtype of M-RV cells and whether they can establish synaptic communication in the brain have remained unknown. M-RV cells engineered to express green fluorescent protein (M-RV-GFP) had functional ion channels but did not establish synaptic communication in vitro. However, when transplanted into newborn mice cerebella, a site of extensive postnatal neurogenesis, these cells expressed endogenous cerebellar granule precursors and neuron proteins, such as transient axonal glycoprotein-1, neurofilament, type-III β-tubulin, superior cervical ganglia-clone 10, glutamate receptor-2, and glutamate decarboxylase. Importantly, they exhibited action potentials and were capable of receiving glutamatergic synaptic input, similar to the native cerebellar granule neurons. These results suggest that M-RV-GFP cells differentiate into glutamatergic neurons, an important neuronal subtype, in the postnatal cerebellar milieu. Our findings suggest that although activation of REST-target genes can reprogram myoblasts to assume a general neuronal phenotype, the subtype specificity may then be directed by the brain microenvironment. STEM CELLS 2010;28:1839–1847
The manipulation of cellular plasticity to repair and restore damaged or diseased neurons has been an area of tremendous research interest over the past few years. Several studies have used neural stem cells (NSCs) or even non-neural stem or progenitor cells such as embryonic stem cells and adult bone marrow stem cells to generate the neuronal phenotype [1–15]. Recent studies have demonstrated transcription factor-induced reprogramming of adult somatic cells into pluripotent stem cells, further advancing this field of research [1, 16–23]. These results suggest that mammalian cells have more potential for plasticity than originally thought [24, 25]. However, the ability of heterologous stem and progenitor cells to differentiate into neurons of a specific subtype has been less well-studied . Only a small fraction of cells are converted into neural cells; thus, detailed molecular studies are difficult to perform.
The transcriptional repressor REST is a negative regulator of neurogenesis [24, 25, 27–29]. Consistent with this role, it is expressed mostly in neural progenitors and non-neuronal cells and rarely in differentiated neurons [24, 25, 27–29]. REST binds a 21-bp consensus DNA sequence (RE-1/NRSE) present in the regulatory region of a number of target genes through its DNA-binding domain [30, 31]. This domain is flanked by two repressor domains, RD1 and RD2 [28–32], that interact with several cellular corepressors to repress chromatin at target promoters [24, 25, 27–29, 32–34]. This regulation of any given target gene expression by REST is influenced by the availability of cofactors and its affinity for the RE1 element [35–42].
Although REST function is required to repress the transcription of multiple neuronal differentiation genes, its absence alone is insufficient to activate a subset of target genes. Therefore, in our ongoing studies, we constructed a recombinant transcription factor, REST-VP16, in which both repressor domains of REST were substituted with the VP16 activation domain of the herpes simplex virus [43–47]. REST-VP16 bound to the same target genes as REST, but functioned as an activator instead of a repressor and directly activated REST target genes . Expression of REST-VP16 in neural stem and progenitor cells was sufficient to cause rapid neuronal differentiation . Surprisingly, REST-VP16 expression in myoblasts (M-RV) blocked their differentiation into myotubes in vitro and converted them into cells exhibiting a neuronal phenotype, including expression of neuronal differentiation genes, depolarization-dependent calcium influx, synaptic vesicle recycling, and survival in the presence of mitotic inhibitors . The M-RV cells also survived in the mouse brain and did not form tumors .
The subtype of neurons derived from M-RV cells and their ability to establish synaptic communication in the brain remain to be determined. Here, we found that in vitro, green fluorescent protein (GFP)-tagged M-RV cells (M-RV-GFP) expressed functional sodium, potassium, and calcium channels and exhibited action potentials, although they did not demonstrate synaptic communication. However, when injected into newborn mice cerebellum, a site of extensive postnatal neurogenesis, these cells were converted into neurons that expressed trans axonal glycoprotein-1 (TAG-1), neurofilament M (NF-M), type III beta tubulin (TUBIII), glutamate receptor 2 (GLU R2), glutamate decarboxylase 1 (GAD1), and superior cervical ganglion 10 (SCG10), similar to endogenous granule neurons. M-RV-GFP-derived neurons also had the capacity to receive N-methyl D-aspartate (NMDA) and non-NMDA synaptic inputs, suggesting that they are glutamatergic neurons with potential for synaptic communication.
MATERIALS AND METHODS
The generation of M-RV cells that express the REST-VP16 transgene has been described previously . National Institutes of Health guidelines were followed for all recombinant DNA research. In brief, we transfected a plasmid expressing GFP and selected for neomycin resistance. REST-VP16 transgene maintenance and expression were achieved by adding hygromycin and doxycycline (Roche Applied Sciences, Indianapolis, IA) at final concentrations of 200 and 20 μg/ml, respectively, to the growth medium. M-RV-GFP cells were maintained in DMEM containing 20% fetal bovine serum, 1% antibiotics or antimycotics, and 2% glutamine at 37°C and 5% CO2 tension. Differentiation was induced by seeding cells on a cover glass coated with polylysine (Sigma Aldrich, St. Louis, MO) and laminin (Invitrogen, Carlsbad, CA) in growth medium for 24 hours and then switching to DMEM containing 2% horse serum, 1% antibiotics or antimycotics, and 2% glutamine at 37°C and 5% CO2 tension for 6–37 days.
Whole-Cell Voltage-Clamp Recording
GFP-positive cells with neurite-like extensions were identified under a microscope (Olympus, Inc., Center Valley, PA) with fluorescent illumination. Visualized whole-cell voltage-clamp recordings were obtained from an identified cell with a glass pipette (resistance 3–5 MΩ) filled with a solution containing 126 mM KCl, 10 mM NaCl, 1 mM MgCl2, 11 mM EGTA, 10 mM Hepes, 2 mM ATP, and 0.25 mM GTP, with a pH adjusted to 7.3 with KOH and osmolarity 280–290 mosmol/l. An AxoPatch-1D amplifier and AxoGraph software (Molecular Devices, Sunnyvale, CA) were used for data acquisition and online and offline data analyses. A seal resistance of 2 GΩ or higher and an access resistance of 15 MΩ or lower were considered acceptable. Series resistance was optimally compensated. Access resistance was monitored throughout the experiment. The holding potential was generally −60 mV. Action potentials were evoked by a single pulse of positive current with increasing intensities through the recording pipette. Miniature excitatory postsynaptic currents (EPSCs) were obtained in 60-second epochs in the presence of 1 μM tetrodotoxin (TTX), 1 mM 4-aminopyridine (4-AP), or 10 μM nifedipine applied through the bath solution. Similarly, 10 μM 6-cyano-7-nitroquinoxaline-2,3-dione, an antagonist of non-NMDA glutamate receptors, and 10 μM 2-amino-5-phosphovalerate were also applied through the bath solution for EPSC experiments in cerebellar brain slices. AxoGraph software was used to detect and record spontaneous synaptic events [48, 49].
Cells Injected into Mice Cerebellum and Brain Slice Preparations
Newborn mice were cryoanesthetized, and 1 μl of cold ×1 phosphate-buffered saline containing 20,000 M-RV-GFP cells was injected into the cerebella using a Hamilton syringe with a 26-gauge needle. These studies were carried out in accordance with institutional rules for animal studies. Mice were anesthetized and killed by decapitation. Their skulls were removed, and their brains were placed in physiologic saline (126 mM NaCl, 2.5 mM KCl, 1.2 mM NaH2PO4, 1.2 mM MgCl2, 2.4 mM CaCl2, 11 mM glucose, and 25 mM NaHCO3). The brain block containing the cerebellum was then mounted on a glass cutting support with cyanoacrylate glue. A vibratome (VT1000S, Leica, Bannockburn, IL) was used to make coronal slices, ∼150–200μm thick, in cold physiologic saline. A single slice was then used for electrophysiologic measurements, as described previously [48, 49].
Single Cell Reverse Transcriptase Polymerase Chain Reaction Assays
GFP-positive cells used for patch clamp electrophysiologic studies were sucked into a glass pipette to extract RNA using the RNeasy Plus Mini kit (Qiagen, Valencia, CA). cDNA synthesis was performed using the Sensiscript reverse transcriptase (RT) kit (Qiagen), as per the manufacturer's instructions. The cDNA served as the template for polymerase chain reaction (PCR) amplification of REST-VP16 transgene sequences. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as an internal control.
The primer sequences used were as follows:
GAPDH (Forward), 5′ GTGTTCCTACCCCCAATGTG 3′;
GAPDH (Reverse), 5′ GTCATTGAGAGCAATGCCAG 3′;
RV (Forward), 5′ CAATGGATGTCTCAAAAGTG 3′;
RV (Reverse), 5′ CAACTAGAAGGCACAGTCGAGG 3′;
SCG10 (Forward), 5′ TCCAACCGAAAAATGAGGTC 3′; and
SCG10 (Reverse), 5′ GGCAGGAAGCAGATTACGAG 3′.
cDNA was amplified for 40 cycles under the following conditions: denaturing at 95°C for 1 minute, annealing at 53°C for 30 seconds, and extension at 72°C for 30 seconds. A second PCR was carried out under the same conditions using 1–2 μl of the PCR product as the template. The reaction products were separated by agarose gel electrophoresis.
For quantitative reverse transcriptase polymerase chain reaction (Q-RT-PCR) experiments, SYBR green-PCR master mix (Applied Biosystems, Foster City, CA) was used. The cDNA that was amplified as described earlier was diluted 20-fold and 1 ml was added to the reaction mix. An initial round of amplification was carried out for 40 cycles under the following conditions: denaturing at 95°C for 1 minute, annealing at 65°C for 30 seconds, and extension at 72°C for 30 seconds. A second round of PCR was carried out under the aforementioned conditions using 1 μl of the PCR product as the template. Each reaction was run three times, and melting peak profiles and mean changes in relative fluorescence intensity over time −d(RFU)/dT were calculated using iCycler iQ version 3.0a software. Normalization was performed relative to 18s RNA.
In addition to REST-VP16 and GAPDH primers describe earlier, the following primers for GLU R2 and GAD1 were used:
GLU R2 (Forward): 5′ GGTTACAGGCACCTCCAAAA 3′,
GLU R2 (Reverse): 5′ GGTTTGTGTCGATTGCATTG 3′,
GAD1 (Forward): 5′ CACAAACTCAGCGGCATAGA 3′, and
GAD1 (Reverse): 5′ CTGGAAGAGGTAGCCTGCAC 3′.
Immunofluorescence and Immunohistochemical Analyses of Cerebellar Slices
Paraffin-embedded cerebellar slices from animals that had been injected with M-RV-GFP cells were deparaffinized and blocked for 1 hour with Tris-buffered saline containing 0.05% Tween-20, 0.2% Triton X-100, and 20% normal goat serum at ambient temperature. The slices were then incubated with Tris-buffered saline containing 0.05% Tween-20, 0.2% Triton X-100, 2% normal goat serum, anti-GFP (Clontech, Mountain View, CA; and Abcam), anti-TUB-III/Tuj1 (Covance, Berkeley, CA), and anti-TAG-1 (Chemicon, Temecula, CA) at 4°C overnight. The primary antibody was omitted in control slices. The slices were washed three times for 10 minutes each with Tris-buffered saline containing 0.05% Tween-20, followed by incubation with Alexa-488 (Molecular Probes, Eugene, OR)- or Cy3 (Sigma Aldrich)-conjugated secondary antibodies for 1 hour at ambient temperature in the same buffer used for the primary antibodies. Nonspecifically bound secondary antibodies were removed by washing the slices three times, as described earlier. TOTO-3 was included during the second washing step to stain nuclear DNA. The stained slices were then covered with slow-fade anti-fade solution (Molecular Probes), and fluorescence was visualized using a Nikon TE2000U microscope. Images were processed using Metamorph and deconvolved using Auto Deblur software. Staining with H&E was performed as described previously . Immunohistochemical analyses were performed using a streptavidin–biotin labeled immunoperoxidase technique (ABC-Elite, Vector Laboratories, Burlingame, CA) with diaminobenzidine as the chromogen. All immunohistochemical analyses were performed using an autostainer (Autostainer Plus, Dako Corp., NY). The mouse monoclonal antibodies applied included those against myogenin, MyoD1, and myosin (Dako, Carpintera, CA). Secondary antibody incubation was performed at ambient temperature for 60 minutes. Meyer's hematoxylin was used as a nuclear counterstain.
Activation of REST Target Genes in Myoblasts Induces a Neuronal Phenotype In Vitro
To Vdetermine whether REST-VP16 promotes functional neurogenesis in M-RV cells in vitro, we generated M-RV-GFP cells, placed them under differentiation culture conditions, and evaluated the expression of functional ion channels in differentiated [M-RV-GFP (D)] cells through patch clamp experiments. M-RV-GFP(D) cells were cultured in the absence of doxycycline to induce REST-VP16 expression, as described earlier . They underwent morphologic changes and exhibited neurite-like extensions; these findings were similar to our previous observations . However, these cells could now be identified by their GFP expression using fluorescence microscopy (Fig. 1A). Furthermore, M-RV-GFP(D) cells showed action potentials (Fig. 1B), indicating that they could form excitable cells.
M-RV-GFP (D) Cells Exhibit Properties of Excitable Cells but Do Not Show Synaptic Transmission In Vitro
To evaluate the ion channels that contribute to excitability in M-RV-GFP(D) cells, we recorded sodium, potassium, and calcium currents under whole-cell voltage-clamp configuration. As shown in Figure 2A, currents were recorded in M-RV-GFP(D) cells that could be abrogated by TTX, an inhibitor of voltage-gated sodium channels. Likewise, currents that could be inhibited by 4-AP were also observed, indicating the expression of potassium channels in these cells (Fig. 2B). Similarly, nifedipine, an inhibitor of L-type calcium channels, attenuated currents in M-RV-GFP(D) cells, demonstrating the presence of functional calcium channels (Fig. 2C). When an action potential was present, it was TTX-dependent, as it was partly mediated by the TTX-sensitive Na+ current and itself was also blocked by TTX (1 μM). Regarding the degree of excitability of these neuron-like cells, a total of 55 cells were examined after cell culture over 30 days. Eight (14.5%) of the total cells had both Na+ and K+ currents and 16 (29.1%) of the total cells had only K+ current (all Na+ current-containing cells had K+ current). Ca2+ currents were also recorded in all Na+ current-containing cells. In normally cultured neurons, every cell had Na+, K+, and Ca2+ currents. It is important to point out that the action potential is apparently different from a typical neuronal action potential trace. This action potential had a lower current overshoot (smaller total action potential amplitude) and a wider duration when compared with a typical neuronal action potential. This could be due to a smaller number of functional Na+ channels and K+ channels acquired, and differential channel kinetics and properties in comparison with those of normal neurons. These results, together with our previous findings  that M-RV(D) cells cultivated in the absence of doxycycline express neuronal differentiation markers (synapsin and type III b-tubulin) and not myogenic markers, suggest that REST-VP16 can reprogram gene expression in myoblasts and induce neurogenesis. These neuron-like M-RV-GFP(D) cells did not exhibit spontaneous EPSCs in culture (data not shown), suggesting that our in vitro system did not support the formation of functional synapses.
M-RV-GFP Cells Transplanted in the Cerebellum Show Properties of Functional Neurons Capable of Receiving Glutamatergic Synaptic Input
Because the cerebellum is a site of substantial postnatal neurogenesis, we determined the ability of M-RV-GFP cells to differentiate into functional neurons in vivo by injecting them into the cerebella of 1-day-old mice . Age-matched control animals were injected with either M-RV-GFP cells or C2C12 myoblasts. The animals were killed after a recovery period of 7 days to prevent extensive migration of transplanted cells from the site of injection and facilitate the identification of GFP-positive cells for the studies described below. Cerebellar slices were prepared and the injection tract and injected cells identified using a fluorescence microscope (Fig. 3A). GFP-positive cells could be detected in these sections as late as 37 days after injection (data not shown). Cerebellar sections from 7-day-old animals were stained with antibodies against neuronal markers, TAG-1, NF-M, TUB-III/Tuj1, or the muscle differentiation markers myosin and myogenin. TAG-1, NF-M, and TUB-III/Tuj1 were expressed in both GFP-positive M-RV-GFP (Fig. 3B) and GFP-negative native cerebellar granule cell neurons (GCN) (Fig. 3C). Under these conditions, control C2C12 myoblasts formed multinucleated myotubes (Fig. 3D, top panel, arrowheads in H&E) and expressed myosin and myogenin (Fig. 3D, bottom panel, arrowheads), suggesting that they had converted to muscle cells, even in the cerebellar microenvironment.
Next, action potentials and synaptic activity were examined in transplanted M-RV-GFP(T) cells in cerebellar slices. In the presence of TTX (1 μM), action potential-independent, spontaneous miniature EPSCs were recorded in some M-RV-GFP cells with a mean frequency of 3.42 ± 1.33 Hz and a mean amplitude of 29.16 ± 3.07 pA (n = 5 cells; Fig. 4A, 4B). These miniature EPSCs were similar to those observed with control cerebellar granule neurons at sites that were distal to the injection tract. These findings indicate that the M-RV-GFP(T) cells had integrated into the cerebellar neuronal network and had formed functional synaptic connections. Further experiments were performed to identify the nature of the neurotransmitter mediating the EPSCs (Fig. 4B). Application of 10 μM 6-cyano-7-nitroquinoxaline-2,3-dione, an antagonist of non-NMDA glutamate receptors, and 10 μM 2-amino-5-phosphovalerate, an antagonist of the NMDA glutamate receptors, abolished the EPSCs of both GFP-positive M-RV-GFP(T) cells and GFP-negative GCNs, suggesting that glutamate was the neurotransmitter mediating the EPSCs [48, 49] observed in both cell types. We also harvested individual cells that were used for patch-clamp studies in glass electrodes and prepared RNA from these cells. Single-cell RT-PCR analysis was performed, using primers designed to amplify transgene sequences. As shown in Figure 4C, the REST-VP16 transgene sequence was amplified in M-RV-GFP(T) cells but not in control GCNs. A product of a similar size was also observed in uninjected control M-RV-GFP(D) cells cultured in vitro. As expected , the REST target SCG10 gene expression was observed in M-RV-GFP(T) and M-RV-GVP(D) cells as well as GCNs. Expression of GAPDH was measured as an internal control. These results were confirmed by SYBR-green quantitative RT-PCR analyses in which REST-VP16, GLU-R2, SCG10, and GAD-1 expression relative to GAPDH was determined (Fig. 4D). These findings confirmed the identity of cells proximal to the injection tract with glutamatergic transmission as injected M-RV-GFP cells expressing REST-VP16. These results suggested that REST-VP16 converted myoblasts into glutamatergic neurons in the mouse cerebellum.
Our data indicate that myoblasts can be reprogrammed to form neuronal cells with glutamatergic properties in the cerebellum by modulating the activity of a single transcription factor, REST. This work has direct relevance to the reprogramming field described earlier [1, 16–23] and suggests that such manipulation through REST-VP16 presents an efficient way to convert other cells into neuronal phenotype. Our observations are consistent with other findings in which cell fate decisions in multipotent progenitors and even differentiated cells have been altered through forced expression of transcription factors. For example, earlier studies showed that the transcription factor MyoD converted fibroblasts into myoblasts . Mesenchymal stem cells were directed toward a neuroectodermal lineage to generate dopaminergic neurons through lentiviral expression of LMX1a [51, 52]. Exogenous Oct4 expression provided neural specification to adipose tissue stromal cells rather than a mesodermal lineage . More recently, in differentiated cells such as fibroblasts, the ectopic expression of a combination of three different transcription factors was sufficient to induce neurogenesis, ion channel expression, and synaptic communication in vitro . In our studies, although M-RV-GFP cells expressed excitable ion channels in vitro, the acquisition of synaptic communication properties required the cerebellar microenvironment. The specific nature of the signal in the cerebellar environment that is necessary to promote functional neurogenesis in M-RV-GFP cells remains to be elucidated, but several studies have highlighted the contribution of cell-extrinsic factors in neuronal differentiation of NSCs. In one study, adult rat hippocampal progenitors required a glial feeder layer to form functional synapses in vitro . However, other studies have shown an important role for glia in the promotion of neurogenesis from NSCs as well as oligodendrocyte precursor cells [35–37, 56–58]. The lack of a feeder layer in our experiments may have contributed to the failure to form synaptic contact in vitro in M-RV-GFP cells.
Intriguingly, M-RV-GFP cells differentiated into glutamatergic neurons in the early postnatal cerebellum. This is the major neuronal subtype in the cerebellum, and granule cell progenitors are known to differentiate into glutamatergic neurons in the postnatal cerebellum. One potential explanation is that the microenvironment, as well as soluble and niche-dependent cell-cell interactions have a role in influencing the nature of the neurons formed . Alternatively, REST-VP16 may exhibit a bias in directing cells toward a glutamatergic fate. These possibilities can be formally tested by injecting M-RV-GFP cells into a different area of the brain, such as the midbrain or forebrain, or into a different anatomic location, such as neuromuscular junctions. In a recent report, the forced expression of Myt1, Mash1, and Brn2 in fibroblasts in vitro gave rise to excitatory glutamatergic neurons, although in vivo studies have found that Mash1 has a prominent role in differentiation into GABAergic neurons [38, 39, 54, 59]. In this context, it would be interesting to determine the subtype fate of fibroblast-derived neurons injected into specific brain regions. The superimposition of extracellular cues on cell-intrinsic signals is further suggested by our finding that control myoblasts differentiated into muscle cells in the cerebellar environment. Thus, cues from the microenvironment alone appeared to be insufficient to reprogram gene expression or alter cell fate decisions in myoblasts in our study. Our findings are in contrast to those of a recent study in which yellow fluorescent protein-labeled cloned myogenic cells formed neurons in the murine brain . The reasons for these differences are not known; however, the brain regions targeted for injection in the two studies were different (cerebellum vs. lateral ventricles). Whether this influenced the differentiation potential of myoblasts is not clear.
Finally, our findings may have therapeutic implications for the treatment of neurodegenerative diseases [40–42, 61–64]. For example, chemical libraries could be screened for small molecules that mimic REST-VP16 function or dsRNA, which converts REST into an activator [33, 42]. Additionally, a more thorough characterization of neurons generated from myoblasts may be feasible since these cells are a more readily accessible source of progenitors compared to ESCs or NSCs.
The data presented here suggest that the sub-type specific differentiation of non-neural progenitors such as myoblasts is not only determined by the activity of specific transcription factors such as REST, but also microenvironment specific cues.
We thank Sei Kameoka for technical help with constructing the M-RV-GFP stable clone and Dr. Raymond Grill for help with our in vivo experiments. This work was supported in part by American Cancer Society Grant RSG-09-273-01DDC to V.G., NIH Grant DA23069 (to Z.Z.P.), and NIH Grants CA81255 and CA97124 (to S.M.). N.D.S.-K. is currently affiliated with the Betty Cowan Research and Innovation Center, Christian Medical College and Hospital, Ludhiana, Punjab, India.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.