Neuronal Transdifferentiation Potential of Human Mesenchymal Stem Cells from Neonatal and Adult Sources by a Small Molecule Cocktail

Human mesenchymal stem cells (MSCs) are good candidates for brain cell replacement strategies and have already been used as adjuvant treatments in neurological disorders. MSCs can be obtained from many different sources, and the present study compares the potential of neuronal transdifferentiation in MSCs from adult and neonatal sources (Wharton's jelly (WhJ), dental pulp (DP), periodontal ligament (PDL), gingival tissue (GT), dermis (SK), placenta (PLAC), and umbilical cord blood (UCB)) with a protocol previously tested in bone marrow- (BM-) MSCs consisting of a cocktail of six small molecules: I-BET151, CHIR99021, forskolin, RepSox, Y-27632, and dbcAMP (ICFRYA). Neuronal morphology and the presence of cells positive for neuronal markers (TUJ1 and MAP2) were considered attributes of neuronal induction. The ICFRYA cocktail did not induce neuronal features in WhJ-MSCs, and these features were only partial in the MSCs from dental tissues, SK-MSCs, and PLAC-MSCs. The best response was found in UCB-MSCs, which was comparable to the response of BM-MSCs. The addition of neurotrophic factors to the ICFRYA cocktail significantly increased the number of cells with complex neuron-like morphology and increased the number of cells positive for mature neuronal markers in BM- and UCB-MSCs. The neuronal cells generated from UCB-MSCs and BM-MSCs showed increased reactivity of the neuronal genes TUJ1, MAP2, NF-H, NCAM, ND1, TAU, ENO2, GABA, and NeuN as well as down- and upregulation of MSC and neuronal genes, respectively. The present study showed marked differences between the MSCs from different sources in response to the transdifferentiation protocol used here. These results may contribute to identifying the best source of MSCs for potential cell replacement therapies.


Introduction
The in vitro generation of neuronal cells from neural (NSCs), embryonic (ESCs), and induced pluripotent stem cells (iPSCs), or by neuronal transdifferentiation of somatic cells by transcription factors (TF) has emerged as a useful strategy for cell replacement therapies in neurological disorders [1][2][3]; however, technical limitations, graft rejection, ethical issues, and/or tumorigenic risk are associated with the neurons derived from such processes [4][5][6]. Therefore, recent efforts have been focused on finding more suitable cell types or avoiding genetic manipulation for the generation of neurons [4,[7][8][9][10][11]. In this respect, mesenchymal stem cells (MSCs) offer some advantages over other cell types. MSCs are potentially able to differentiate into various cell lineages (including neurons), are easy to isolate and expand, have a low tumorigenic risk and low grafting rejection, and lack ethical issues [12][13][14][15]. These properties point to MSCs as suitable sources for cell replacement therapy in neurological disorders [16][17][18][19]; however, an optimal protocol to induce their conversion into neurons remains unestablished.
Chemical compounds known as small molecules have been shown to replace exogenous TF during cell reprogramming [7][8][9]11]. Recent reports demonstrated the neuronal transdifferentiation of fibroblasts and astrocytes by small molecule cocktails [20][21][22][23]. These molecules act by modulating signaling pathways and epigenetic mechanisms implicated in cell reprogramming, neuronal specification, or neuronal survival [21], representing a convenient strategy to avoid the risks of genetic manipulation in the generation of induced neurons. In our previous report, after a small molecule screening assay, we found that a cocktail containing I-BET151, CHIR99021, forskolin, RepSox, Y-27632, and cAMP (ICFRYA) induced the formation of cells with neuron-like morphology and positive for TUJ1 and MAP2 from bone marrow-(BM-) MSCs [10].
MSCs can be isolated from many adult and neonatal tissues. However, comparative studies indicate that the MSCs from different tissues present differences in the efficiency of trilineage differentiation and other functional abilities, even though they meet the properties to be considered MSCs [24][25][26][27]. The present study is aimed at comparing the neuronal transdifferentiation potential of adult and neonatal MSCs obtained from different sources. To this end, we evaluated the neuronal-like morphology and neuronal markers induced by the ICFRYA cocktail in MSCs obtained from bone marrow (BM), skin (SK), dental pulp (DP), periodontal ligament (PDL), gingival tissue (GT), Wharton jelly (WhJ), placenta (PLAC), and umbilical cord blood (UCB). Neuronal induction was successful in the MSCs from some but not all sources. Strategies were selected to improve the induction of the MSC sources that showed neuronal properties. The presence of mature neuron markers, changes in global gene expression, and electrophysiological activity were examined in cells in which neuronal transdifferentiation was presumed.

Reagents and Antibodies.
Neurobasal medium, α-MEM, RPMI medium, DMEM low glucose (DMEM-lg), DMEM/F-12, fetal bovine serum (FBS), L-glutamine, dispase II, Glutamax, antibiotics, trypsin/EDTA, human neurotrophic factors, N2, and B27 were all purchased from Gibco™ Thermo Fisher Scientific Inc. (MA, USA). Fibronectin, heparin, gelatin, nonessential amino acids, Hoechst 33258, and normal goat serum (NGT) were acquired from Sigma-Aldrich, Merck (St. Louis, MO, USA). The following primary antibodies were used: mouse anti-TUJ1 (SC-80016) and rabbit anti-MAP2 (SC-20172) obtained from Santa Cruz Biotechnology (CA, USA); rabbit anti-GABAB (SC-376282), ENO2 (GTX113428), TAU (GTX116044), and mouse anti-NF-H (GTX27795) from GeneTex Inc. (CA, USA); and rabbit anti-NeuN (MAB377) from Merck-Millipore (Darmstadt, Germany Mononuclear cells were isolated from bone marrow and umbilical cord blood samples by a Ficoll gradient, and cells from the internal area of the central placenta lobules were obtained using an enzymatic digestion procedure (Trypsin-EDTA). These cell samples were cultured in DMEM-lg plus 4 mM L-glutamine, 50 U/mL of penicillin, 50 μg/mL of streptomycin, and 50 μg/mL of gentamicin (DMEM-lg) and supplemented with FBS-10% [28]. The Wharton's jelly (membrane), gingival tissue, periodontal ligament, and dental pulp tissue samples were cut into small pieces and cultured as explants in α-MEM supplemented with FBS-15%. Finally, the skin samples were placed overnight in RPMI medium and dispase II, and then, the dermis was mechanically separated from the epidermis. Cells from the dermis were isolated by allowing them to migrate from the dermal segments placed in culture (DMEM-lg supplemented with FBS-10%). For all samples, the medium was changed every two days. The adherent cells were subcultured using trypsin/EDTA at 2 × 10 2 cells/cm 2 in DMEM-lg supplemented with FBS-10%. All experiments were performed with MSCs between the fourth and sixth passages.

2.4.
In Situ Analysis of Cell Viability. Cells were incubated for 10 min with 1 μg/mL propidium iodide. Microphotographs were obtained using a direct epifluorescence microscope Olympus 1X71 with 10x magnification and the QCapture Pro 6.0 software. Then, the cells were incubated for 1 h with goat anti-mouse Alexa 488 or goat anti-rabbit Alexa 568 as appropriate. The nuclei were counterstained with 1 μg/mL Hoechst 33258 diluted in PBS. Microphotographs were obtained by the direct epifluorescence microscope Olympus IX71 using the QCapture Pro 6.0 software, and data were analyzed with ImageJ software. At least 500 nuclei from five randomly selected fields were counted to calculate the percentage of positive cells.
2.6. Real-Time Quantitative PCR (qPCR). RNA samples from cultures were extracted with the GenElute Mammalian Total RNA Miniprep Kit (RTN70 Sigma-Aldrich, St. Louis, MO, USA) according to the manufacturer's instructions. The concentration of RNA samples was quantified by a NanoDrop spectrophotometer (Thermo Scientific). Complementary DNA (cDNA) was synthesized from 1 μg RNA with the iScript cDNA Synthesis Kit (Bio-Rad Laboratories Inc., CA, USA) and a Veriti® 96-Well Thermal Cycler (Applied Biosystems). PCR primers were designed using the last reference sequence (RefSeq) version of each gene with the PrimerBlast software; Supplementary Table 5 includes a list of primers used in this study. The qPCR reactions were conducted using a QuantiFast SYBR Green PCR Kit (Qiagen, MD, USA), according to the manufacturer's instructions, and run on a StepOnePlus Real-Time PCR system (Applied Biosystems). After amplification, the melting curves of the RT-PCR products were determined to demonstrate product specificity. PCR efficiency was optimal and ranged from 90% to 100% in the different target gene qPCR assays. For the selection of the best reference gene, a comparison of the transcriptional variation of different housekeeping genes in response to the conditions was performed. The gene with the least CT variation between treatments and control cultures was selected. Thus, gene expression was normalized to TATA-Box binding protein (TBP) mRNA. The relative quantification of mRNA levels was performed using the Pfaffl method [30].

Microarray
Assay. Cells were incubated as described in the neuronal maturation protocol. The samples were then processed for RNA extraction. Total RNA (10 μg) was used for cDNA synthesis incorporating dUTP-Alexa 555 or dUTP-Alexa 647 and employing the CyScribe First-Strand cDNA labeling kit (Amersham). Equal quantities of labeled cDNA were hybridized using the hybridization solution HybIT2 (TeleChem International Inc.) on the collection of human arrays (10 K) for 14 h at 42°C. The acquisition and quantification of array images were performed in the ScanArray 4000 with its accompanying software from Packard BioChips. For each spot, the mean density, mean background, and mean normalized signal values were calculated with the Array-Pro Analyzer software from Media Cybernetics. Microarray data analysis was performed with the free software genArise, which was developed in the Computing Unit of the UNAM Cellular Physiology Institute (http:// www.ifc.unam.mx/genarise/). The microarray data were deposited in the Gene Expression Omnibus database repository (GSE120681). The functional annotation gene analysis analyzed the up-and downregulated genes derived from the microarray assay using DAVID 2.0 (https://david .ncifcrf.gov/tools.jsp) and bioinformatics resources to obtain the Gene Ontology functional annotation classification of the genes.
2.9. Statistical Analysis. The results were expressed as the mean ± standard error (SEM) from three independent experiments of three biological samples or 4-6 independent experiments from one biological sample as indicated. The statistical analyses were performed using Student's t-test for between-group comparisons and a one-way ANOVA and the Tukey test for multiple comparisons. GraphPad Prism software for Windows version 6 (La Jolla, CA, USA) was used for all statistical procedures. Differences with p < 0 05 were considered statistically significant.  Table 1) and characterized according to the criteria defining human MSCs proposed by the International Society for Cellular Therapy [31]. For all MSC samples, the cells adhered to the plastic exhibited fibroblastic morphology (Supplementary Figure 1), were positive for MSC markers (CD105, CD90, and CD73), were negative for hematopoietic markers and HLA-DR surface molecules (Supplementary Table 2), and were capable of differentiating into adipocytes, osteoblasts, and chondrocytes (Supplementary Table 3). The neuronal medium per se (NM: N2/B27/hFGF) did not induce neuronal morphology or reactivity to neuronal markers (Supplementary Figure 3); however, a weak reactivity against TUJ1 was observed in all MSC sources (Figures 1(b

Neuronal Induction by the ICFRYA Cocktail of Adult and
Neonatal MSC Sources. Neuronal induction of all MSC sources was conducted using the ICFRYA small molecule cocktail protocol (Figure 1(a)) as previously reported for BM-MSCs [10]. In addition to neuronal properties, necrotic cell death was evaluated by propidium iodide staining. The neuronal induction effects of the ICFRYA cocktail on MSCs from different sources varied. Neuron-like cells with reactivity to TUJ1 were induced in DP-MSCs (50 ± 7 7%), PDL-MSCs (42 ± 7 1%), GT-MSCs (32 ± 1 9%), SK-MSCs (34 ± 9 1%), and PLAC-MSCs (67 ± 2 3%); however, very low percentages of MAP2 + cells were found (0.7% to 5.9%) (Figures 1(b)-1(f), right pictures and graphs). In these MSC sources, the cocktail produced 48%-68% necrosis. In WhJ-MSCs, the chemical cocktail did not evoke cells with neuronal properties, but high levels of necrosis were present (Figure 1(g) right picture and graph). In UCB-MSCs, the ICFRYA cocktail induced neuronal properties, i.e., bipolar morphology with a high percentage of TUJ1 + cells (57 ± 3 3%) and coexpression of MAP2 in 27 ± 3 9%; however, a very high rate of necrosis (84 ± 6 8%) was observed (Figure 1 Table 4). It is noteworthy that I-BET alone did not exhibit any effects on the MSCs from any tissue (data not shown). I-BET151 and forskolin are considered key molecules for chemical-based neuronal transdifferentiation; in other reports, high concentrations of these molecules have been used for neuronal transdifferentiation [22]. However, increasing the concentration of I-BET151 (1 μM to 2 μM) and/or forskolin (50 μM to 75μM) did not improve the expression of neuronal properties on WhJ-, DP-, PDL-, GT-, SK-, or PLAC-MSCs but did increase necrosis (Supplementary Table 4 and Supplementary Figure 2). As illustrated in Figures 2(a) and 2(e), very few cells from WhJ-MSCs presented a bipolar morphology and reactivity to TUJ1. In PDL-MSCs, although MAP2 reactivity was increased, the signal was not associated with cell protrusions, and almost all cells maintained the bipolar morphology (Figures 2(b) and (e)). On the other hand, in UCB-MSCs incubated with ICFRYA containing 2 μM I-BET151, neuronal morphology was observed after two days of induction (Figure 2 (Figure 4(a)). The fold change in expression of CD90, CD105, DCX, and ND1 was significantly higher in the cultures generated from UCB-MSCs than from BM-MSC cells.
To evaluate the transdifferentiation process, we conducted a global RNA expression analysis. In the presence of the ICFRYA cocktail, both UCB-and BM-MSC cells upregulated the expression of neuronal regulators and other genes associated with neurodevelopment and functional properties ( Figure 4(b)). Among the genes implicated in functional differentiation, the potassium voltage-gated gene KCNC1 was upregulated in BMs, and the KCNC1 and KCNJ16 genes were upregulated in UCB-MSCs. The glial genes GFAP and S100B remained unchanged, and positive regulators of the cell cycle and genes associated with mesenchymal and fibroblastic lineages were downregulated (Figure 4(b)). Gene Ontology term analysis indicated enrichment in the genes related to processes such as small molecule metabolism, cell signal, cell differentiation, positive regulation of neurogenesis, neuronal differentiation, functional neuron activities, dendritic spine development, and action potentials (Figure 4(c)). Altogether, the results indicate that the neuron-like cells obtained can be considered induced neurons (iNs) generated from UCB-MSCs (UCB-iNs) or iNs generated from BM-MSCs (BM-iNs). Although the UCB-iNs and BM-iNs have similar neuronal features, there was a higher number of genes regulated in each Gene Ontology process and a higher p value in the UCB-iNs (Figure 4(c)). 5(b)). Fast inward sodium-like currents were not detected in the control nor treated conditions. Outward noninactivating ion potassium-like currents were also evaluated. Slight and irregular potassium-like currents were recorded in BM-iNs and no different to those in control BM-MSCs (Figures 5(c) and 5(d)). Potassium-like currents of high    MAP1B  MAP1LC3B  B2M  TBCB  ENO3  DCX  GABARAPL2  GABRA3  GRM2  PVALB  NEU1  RIT2  BDNF  NDNF  NLGN1  GCNT3  TGIF1  IFNG  NREP  RAC1  SNX3  IFT140  NTRK2  RHEB  CERCAM  RANBP2  FGF12  NEUROD2  BRD2  BRD7  WNT7A  WNT2B  RHOA  ACSL3  FXYD2  CLIC2  CHPT1  KCNC1  KCNJ16  SNAP29  GFAP  S100B  CD14  CD22  CD53  CD28  FGF20  ENG  ZER1  UTF1  GRB7  BCAR1  KITLG  CSE1L  MKl67  CCNC  DOCK4  MAP3K7  CDC42

Discussion
Neuronal transdifferentiation without genetic manipulation from a suitable cell source is the most relevant cell replacement strategy in the brain. This strategy was approached in the present study by generating neuron-like cells from human MSCs with a small molecule cocktail.
The neuronal induction potential of neonatal and adult MSC sources was compared since some studies suggest that MSCs from neonatal sources may have greater plasticity and stemness than adult sources, possibly due to their lower degree of commitment [24,32,33]. This hypothesis was not confirmed in the protocol using the ICFRYA small molecule cocktail as the neuronal inductor because the best sources to generate cells with neuronal features were BM-MSCs and UCB-MSCs. In contrast, neuronal traits were not induced from WhJ-MSCs or were limited in the MSCs from dental tissues, SK-MSCs, and PLAC-MSCs.
In addition to the neuronal transdifferentiation potential, other studies have shown that MSCs from different tissues exhibit differences in other functional activities, such as cell proliferation, immune regulation properties, and trilineage differentiation potential [12,26,27]. The intrinsic properties of each MSC source or their tissue-specified functions may explain the different responses; however, this has not been demonstrated. Each of the small molecules in the ICFRYA cocktail used to induce neuronal transdifferentiation regulates specific signaling pathways or epigenetic mechanisms [7,9,11,34,35]. By performing screening assays with small molecule combinations, it is possible to define whether the differences in the neuronal induction efficiency of each specific cocktail are related to the particular signaling pathway or epigenetic mechanism affected by the cocktail molecules in the different MSC sources. Another relevant feature to be considered in the process of neuronal transdifferentiation is that cell death may function as a regulatory strategy in vivo to accomplish successful neurogenesis and neuronal connection in the neural network [36,37]. During in vitro neuronal transdifferentiation, cell death could be a destabilization product of the process itself (activation/inhibition of metabolic pathways), resulting in the elimination of cells that could not adapt to changes. In other studies of neuronal transdifferentiation with small molecule cocktail protocols, cell viability was not examined or presented as a marker of induction efficiency [20,21]. Here, we evaluated a possible correlation between necrosis and neuronal induction efficiency and found that the ICFRYA cocktail evoked high cell death without neuronal induction in WhJ-MSCs and high neuronal induction efficiency with low cell death in BM-and UCB-MSCs (Figure 2 and Supplement Table 4).
The relevant role of I-BET151 in the neuronal induction by small molecule cocktails has been reported for mouse fibroblasts [22], human astrocytes [20], and in our previous study on BM-MSCs [10]. This effect was extended for all the MSC sources studied here, since removing I-BET151 from the cocktail blocked the presence of neuronal properties regardless of the tissue source. I-BET151 is a regulator that disrupts the epigenetic memory of the original cell, thus favoring cell reprogramming [34,[38][39][40]. I-BET151 also arrests the cell cycle, a process that is associated with neuronal transdifferentiation [34,41]; however, I-BET151 alone did not induce neuronal properties. The ICFRYA cocktail contains forskolin and dbcAMP, which increase the levels and synthesis of intracellular cAMP. CHIR99021 inhibits GSK3/beta catenin and promotes the expression of neuronal genes. RepSox is involved in the reprogramming processes that inhibit SMADS, and Y27632 inhibits ROCK to maintain cell viability [7,35,42]. Altogether, this suggests that the process of neuronal transdifferentiation requires the synergic regulation of pathways involved in neuronal transdifferentiation, neuronal specification, and survival.
During direct transdifferentiation, the original cells must first undergo partial dedifferentiation, allowing the repression of the lineage of origin and then differentiate into a new cell type [43,44]; moreover, the cells have to arrest the cell cycle at the same time that the chromatin is modulated [41]. In BM-MSCs and UCB-MSCs, the ICFRYA cocktail enriched with neurotrophic factors (NT3, BDNF, and GDNF) favors the generation of cells with complex neuronal morphology, reactivity to mature neuronal markers, and upregulated neuronal genes and neuronal process. Moreover, mesenchymal markers were decreased, and genes related to mitosis, cell cycle, and mesenchymal lineage were downregulated. Considering all these criteria, the transdifferentiated cells were determined to be induced neurons generated from BM-MSCs (BM-iNs) and UCB-MSCs (UCB-iNs).
To be considered a functional iN, cells should exhibit electrophysiological activity, such as action potentials and synaptic transmission. To examine functional properties, the presence of voltage-gated potassium and sodium currents was evaluated in BM-iNs and UCB-iNs. Sodium currents were not detected, and potassium-like currents were present only in UCB-iNs. These results suggest that UCB-iNs may be in a transition phase towards functionally induced neurons. To obtain functional neurons, it is necessary to test other maturation strategies. Three-dimensional (3D) cultures have been shown to be an option to obtain functionally mature cells during the reprogramming or transdifferentiation processes of several cell types [45]. This option, together with the ICFRYA cocktail, may be a convenient strategy to generate functional iNs from MSCs.
It has been recognized that MSCs present advantages over other cell types and have a low tumorigenic risk in cell replacement therapies. Moreover, clinical studies in human and neurodegenerative animal models indicate that MSCs improve regeneration capacity and ameliorate brain injuries. Although BM-MSCs are commonly considered the "gold standard," other MSC sources are more accessible and raise fewer ethical issues. Therefore, identifying the most suitable cell type and cell source for neuronal conversion is an essential matter in cell therapy approaches for neurological disorders.

Conclusion
This study determines that human MSCs can be converted into transgene-free neuronal cells and establishes UCB-MSCs as the most suitable sources for undergoing neuronal transdifferentiation by a specific small molecule cocktail.

Data Availability
The microarray data used to support the findings of this study have been deposited in the GEO repository (GSE120681).  Table 1, Supplementary Table 2, and Supplementary Table 3