Transdifferentiation of periodontal ligament-derived stem cells into retinal ganglion-like cells and its microRNA signature

Retinal diseases are the leading causes of irreversible visual impairment and blindness in the developed countries. Human retina has limited regenerative power to replace cell loss. Stem cell replacement therapy has been proposed as a viable option. Previously, we have induced human adult periodontal ligament stem cells (PDLSCs) to the retinal lineage. In this study, we modified our induction protocol to direct human adult PDLSCs into retinal ganglion-like cells and determined the microRNA (miRNA) signature of this transdifferentiation process. The differentiated PDLSCs demonstrated the characteristics of functional neurons as they expressed neuronal and retinal ganglion cell markers (ATOH7, POU4F2, β-III tubulin, MAP2, TAU, NEUROD1 and SIX3), formed synapses and showed glutamate-induced calcium responses as well as spontaneous electrical activities. The global miRNA expression profiling identified 44 upregulated and 27 downregulated human miRNAs after retinal induction. Gene ontology analysis of the predicted miRNA target genes confirmed the transdifferentiation is closely related to neuronal differentiation processes. Furthermore, the expressions of 2 miRNA-targeted candidates, VEGF and PTEN, were significantly upregulated during the induction process. This study identified the transdifferentiation process of human adult stem cells into retinal ganglion-like cells and revealed the involvement of both genetic and miRNA regulatory mechanisms.

The microRNA signatures of retinal-induced human PDLSCs. To understand the molecular mechanism of retinal induction effect, global miRNA expression profile (miRNome) of retinal-induced PDLSC was determined by microarray. The miRNA expression profiles of groups at Day 0 and Day 24 were differentially clustered and separated from each other by either hierarchical clustering (Fig. 5A) or principle component analysis (Fig. 5B). A total of 170 miRNAs were expressed at a two-fold difference in treated PDLSC at Day 24 compared to Day 0. Among these 170 miRNAs, 71 of them were differentially expressed (p corr < 0.05 and fold change > 2; Table 1). 44 of them were upregulated, and 27 of them were downregulated. Based on expression levels, the predicted miRNA target genes and reported neuron-related miRNA 17,18 , 5 miRNAs (hsa-miR-132, hsa-miR-29b, hsa-miR-30d, hsa-miR-630 and hsa-miR-7) were selected for validation. hsa-miR-29b and hsa-miR-7 were downregulated throughout the treatment period (Fig. 6). In contrast, hsa-miR-132 and hsa-miR-630 were upregulated along the retinal induction treatment. These 4 miRNAs showed significant difference during the treatment period when compared to Day 0. However, hsa-miR-30d did not show significant fold-change differences. Therefore, 4 out of 5 selected miRNAs were validated. Notably, comparing to the reported miR-NAs in eye and retinal progenitor development, RPE differentiation and retinal degeneration [13][14][15][16]19,20 , 12 miRNAs showed p corr < 0.05 (hsa-let-7b * , hsa-let-7d, hsa-let-7e, hsa-let-7f-1 * , hsa-let-7i, hsa-miR-125a-3p, hsa-miR-125a-5p, hsa-miR-15b, hsa-miR-181a, hsa-miR-18a, hsa-miR-214, and hsa-miR-301a; Table 2). Subsequently, a global target gene list of the retinal-induced miRNAs (2043 genes) was generated by the TargetScan in the GeneSpring (Agilent) platform. Gene ontology of this gene list, analyzed by DAVID (1942 DAVID identities), revealed that retinal-induced miRNAs might target the genes involved in different processes of neuron differentiation (Table 3). microRNA-132 in retinal progenitor cell differentiation and its target gene analysis in human PDLSCs. In order to delineate the biological roles of the differentially expressed miRNAs in retinal differentiation process, retinal progenitor cells from embryonic day 18 rats were transfected with an expression plasmid inserted with miR-132 precursor sequence and treated with differentiation medium. Compared to the empty vector control (pCMVMIR), downregulation of Pax6 gene (1.5 folds, p < 0.001) and upregulation of Brn3a and Brn3b genes (4.2 folds, p < 0.01 and 5.1 folds, p < 0.001, respectively) were observed in retinal progenitor cells transfected with miR-132 expression plasmid (Fig. 8A). Rho and Gfap genes showed no significant difference between empty vector control and miR-132-transfected cells. The retinal progenitor cell differentiation analysis suggested that miR-132 could direct retinal progenitor cells towards RGC lineage.

Discussion
Stem cell therapy for retinal diseases has entered the stage of clinical trials. Autologous bone marrow-derived mononuclear cells have been applied in patients with retinitis pigmentosa or cone-rod dystrophy 23 . Moreover, human ESC-derived retinal pigment epithelial cells were used to treat patients with Stargardt's macular dystrophy and advanced non-exudative age-related macular degeneration by cell replacement therapy 24 . But adult stem cell-derived retinal cells have not been attempted for use in cell replacement therapy. We aim to study human adult stem cells as they can be conveniently obtained from patients for autologous transplantation or in vitro drug screening, which is the basis of personalized medicine in future. We made a progress in this direction when we induced human adult PDLSCs to   retinal fate 11 . Results of this present study confirm the feasibility of retinal induction on human PDLSCs. We found PDLSCs underwent the noggin-Dkk1-IGF1 induction treatment not only produced photoreceptor cells as we have previously reported 11 , but also expressed RGC and neuronal markers (ATOH7, POU4F2, β -III tubulin, MAP2 and TAU; Figs 1 and 2), formed synapses, gave glutamate-induced calcium responses (Fig. 3) and demonstrated spontaneous electrical activities ( Fig. 4) in this study. The mechanism of this retinal induction process warrants further investigation. This study also, for the first time, reported the miRNA signatures of retinal induction on human adult stem cells. After 24-day retinal induction treatment, 71 human miRNAs were differentially expressed, which 44 of them were upregulated and 27 were downregulated (Table 1). Comparing to the reported miRNAs related to eye and retinal progenitor development, RPE differentiation and retinal degeneration [13][14][15][16]19,20 , 12 miRNAs showed p corr < 0.05 ( Table 2). Six of them (hsa-let-7f-1, hsa-let-7i, hsa-miR-125a, hsa-miR-15b, hsa-miR-18a and hsa-miR-301a) were commonly found in the differentiation treatment of ESC into retinal pigment epithelial cells 13 . Moreover, let-7 and miR-125 families are key regulators of retinal progenitor cell development 16 . In addition, 17 miRNAs (miR-136, miR-143, miR-148a, miR-15b, miR-18a, miR-181a, miR-181a*, miR-20b, miR-27b, miR-29b, miR-30d, miR-30e*, miR-301a, miR-376a, miR-376b, miR-410 and miR-7), which are differentially expressed in our retinal induction treatment, are involved in the regulation of developing mouse retina 25 . Thus our miRNA profile is closely related to the retinal cells formation processes, indicating that the commonly found miRNAs may guide human PDLSCs to retinal precursors with a similar differentiation pathway as ESC and retinal development in mice.
Among the upregulated miRNAs, miR-132 is expressed in RGCs 26 and regulates neuron development and neurite morphology 17,18 . In this study, we, for the first time, showed that miR-132 could be involved in the RGC differentiation from retinal progenitor cells as its overexpression in retinal progenitor cells upregulates mature RGC markers (Brn3a and Brn3b) expression (Fig. 8A). This further confirms that miR-132 contributes to the differentiation of human PDLSCs into retinal ganglion-like cells. Moreover, we also demonstrated that miR-132 could downregulate CUX1 protein expression in human PDLSCs (Fig. 8C). Since reduction in Cux1 expression has been shown to promote neurite outgrowth in cortical neurons 27 , the morphology of human PDLSC-derived neurons could be related to the interactive functions of miR-132 and CUX1 in our trans-differentiation process. In addition, the properties of miR-132 are  Continued consistent with the gene ontology analysis of the predicted miRNA target genes (Table 3). Furthermore, BDNF and bFGF were suggested to induce miR-132 expression 28,29 , indicating that BDNF and bFGF in our treatment medium might be responsible for the upregulation of miR-132 in the induced PDLSCs. Notably, the highest upregulated miRNA, hsa-miR-630, was correlated with cancer cell death 30,31 . This is the first report that it could be associated with neuronal differentiation of human adult stem cells. In addition, GJA1, which is a negative modulator of neuronal differentiation 32 , was downregulated along the retinal induction period (Fig. 7), further affirming our PDLSC proceed towards neuronal differentiation under the retinal induction treatment. We chose the 5 miRNAs (hsa-miR-132, hsa-miR-29b, hsa-miR-30d, hsa-miR-630 and hsa-miR-7) for validation not just based on the previous reports or the expression levels, but also the predicted target genes. VEGFA is the target gene of hsa-miR-29b and hsa-miR-15b (Supplementary Table 2), which were downregulated during the retinal induction treatment (Table 1 and Fig. 6). From miRNA target expression analyses, we found that VEGFA gene expression, VEGF protein expression and VEGF 165 secretion were all increased (Fig. 7). These findings verify the prediction that VEGFA upregulation could be correlated with hsa-miR-29b and hsa-miR-15b downregulation. Furthermore, hypoxia, which upregulates VEGF, has been shown to increase retinogenesis from ESC 33 . Therefore, the elevated VEGF level, together with downregulation of hsa-miR-29b and hsa-miR-15b, should account for retinal differentiation of human adult stem cells.
Our retinal induction protocol produces retinal ganglion-like cells (Figs 1-4). The selected target genes (KLF4, MAP3K12, PTEN and SOCS3) are closely related to RGCs for gene expression analysis 21,22,34 . MAP3K12, PTEN and SOCS3 genes as well as the PTEN protein were upregulated during the retinal induction process (Fig. 7). Since deletion of Pten at the onset of neurogenesis in retinal progenitor cells results in the reduction of RGCs and rod photoreceptors 35 , our results further suggest that the treated PDLSC undergo proper retinal/neuronal differentiation guided by PTEN regulation. Moreover, KLF4, the predicted target of hsa-miR-7 (Supplementary Table 3), was upregulated during the retinal induction treatment (Fig. 7). Accordingly, KLF4 could also participate in RGC development during PDLSC retinal differentiation 34 . Nevertheless, we predicted that KLF4 upregulation could be correlated with hsa-miR-7 downregulation (Figs 6 and 7), which is confirmed by a recent study on neural stem cell differentiation 36 . In-depth functional analyses as well as the relationship with putative target genes are needed to validate the role of miRNAs in retinal induction process of human adult stem cells.
In summary, this study, for the first time, demonstrated transdifferentiation of human PDLSCs into functional retinal ganglion-like cells, and identified miR-132, VEGF and PTEN as key regulators in the retinal fate determination of human adult stem cells. Results from this study reveal the involvements of both genetics and miRNA regulatory mechanisms in human adult stem cell retinal differentiation.

Methods
Human periodontal ligament-derived stem cell culture. Human PDLSC lines were established previously 10,11 . These cells were cultured in Dulbecco's modified Eagle's medium (high glucose; Gibco BRL, Rockville, MD) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Gibco BRL). Cells with passage 3-5 were used for the retinal induction experiments.

Glutamate-evoked calcium response. Spontaneous intracellular calcium transient was evaluated
using fluo-4-acetoxymethyl ester (Fluo-4AM; Invitrogen) 11 . Briefly, the treated or control PDLSCs at Day 24 were incubated in Hanks' balanced salt solution (HBSS, Ca 2+ /Mg 2+ -free; Gibco BRL) containing Total RNA was collected at Day 0, 10, 17 and 24. Five significant miRNAs from the microarray profile (hsa-miR-132, hsa-miR-29b, hsa-miR-30d, hsa-miR-630 and hsa-miR-7) were validated using TaqMan PCR approach. snRNA U6 was used for normalization. The relative fold changes were compared to the group at Day 0. The data represented, mean ± standard deviation. *p < 0.05, **p < 0.01, ***p < 0.001.  Electrical activity analysis. Spontaneous electrical activities of the induced PDLSCs were detected by a microelectrode array (MEA) system (Multi Channel Systems, Reutlingen, Germany) on a laminin-coated 54-electrode 6-well MEA chip at Day 24 of differentiation. The differentiated cells were seeded onto the MEA chip at least 7 days prior to recording. The MEA signals, sampled at 25 mHz for 200 seconds, were amplified and digitized using the MEA-2100-60 system (Multi Channel Systems) with integrated temperature controller, data acquisition interface. The spike events were then filtered and analyzed using the MCRack software (Multi Channel Systems). The field potentials where electrodes exhibiting spike frequencies below 0.1 Hz were removed by a 0.1 Hz High Pass filter based on reported minimal steady firing rates 39 . Spikes were detected after spike sorting at a threshold of 15 μ V 40 . The MEA signal of purified neonatal RGCs was used as a reference.
microRNA microarray and data analysis. The protocol for microRNA microarray analysis has been established previously 41 . Briefly, total RNA, including the miRNA fraction, in TRIzol reagent was  Downstream mRNA targets of the miRNAs were predicted by TargetScan (http://genes.mit.edu/targetscan/index.html) in the GeneSpring GX 11.5 software (Agilent). Context percentile of 95 was used as the criteria for target prediction. Gene enrichment and gene ontology analysis was performed by DAVID Bioinformatics Resources 6.7 (http://david.abcc.ncifcrf.gov/) 42 . Enrichment score greater than 1.3 was considered as significant. Based on expression levels, the predicted miRNA target genes and reported neuron-related miRNA 17,18 , 5 miRNAs from microarray results (hsa-miR-132, hsa-miR-29b, hsa-miR-30d, hsa-miR-630 and hsa-miR-7) were selected for validation. Total RNA (20 ng) was reverse transcribed using the TaqMan MicroRNA Reverse Transcriptase kit (Applied Biosystems, Forster City, CA). The resultant products were quantified using the appropriate TaqMan MicroRNA Assays (Applied Biosystems) on a Stratagene Mx3005P Real-Time PCR Detection System (Stratagene, La Jolla, CA). Results were all normalized to U6 expression. Independent T-test was used for statistical analysis. Three independent samples from each time-point were used in the validation experiment.
For gene expression analysis of the predicted miRNA target genes, Sybr green PCR (Applied Biosystems) was performed on a real-time PCR machine (Stratagene, La Jolla, CA). Housekeeping gene (GAPDH) was used for normalization. The relative expression levels were compared to that of Day 0.
For immunoblotting analysis, the induced PDLSCs were lysed by RIPA buffer supplemented with protease and phosphatase inhibitors. The total protein concentrations of the cell lysates were measured by Protein assay (BioRad, Hercules, CA). Equal amount of total protein (10 μ g) for each denatured samples were resolved on 12.5% SDS-polyacrylamide gel and electro-transferred to nitrocellulose membranes for sequential probing with the primary antibodies and secondary antibodies conjugated with horseradish peroxidase (Santa Cruz Biotechnology, Santa Cruz, CA). The signals were detected by enhanced chemiluminescence (Amersham Pharmacia, Cleveland, OH) with the ChemiDoc TM XRS + system (BioRad). β -actin was used as housekeeping protein for normalization.
For immunofluorescence analysis, the induced PDLSCs were fixed in 4% paraformaldehyde (Sigma Aldrich, St. Louis, MO). After permeation and blocking, the cells were labeled with different primary antibodies against RGC or neuronal markers and secondary antibodies conjugated with Alexa Fluor ® 488 or 594 (Santa Cruz Biotechnology). DAPI was used for the nuclear staining. The fluorescence signals were visualized under a fluorescence microscope (Eclipse Ni-U; Nikon).
VEGF 165 secretion was analyzed by ELISA assay. Briefly, culture supernatant was concentrated by a 3-kDa centrifugal filtration unit (Millipore, Billerica, MA), and total protein concentration was measured by Protein assay (BioRad). Equal amount of total protein (1 ng) was applied to the human VEGF Quantikine ® ELISA assay (R&D Systems Inc., Minneapolis, MN) according to the manufacturer's  TargetScan-predicted miR-132 target genes related to neuron differentiation and negative regulation of cell differentiation. (C) Expression analysis of CUX1 protein on the miR-132-transfected human PDLSCs by immunoblotting. β -actin was used as housekeeping protein for normalization, and the pCMVMIRtransfected cells were used as control. **p < 0.01, ***p < 0.001.
protocol. Absorbance at 450 nm with a reference of 540 nm was measured by a spectrophotometer (Powerwave XS, Bio-Tek Instruments). The amount of VEGF 165 secreted to the culture supernatant (pg of VEGF 165 /ng of total protein in culture supernatant) were then determined.
Retinal progenitor cell differentiation. Retinas from embryonic day 18 Sprague Dawley rats were digested with 0.05% trypsin (Gibco BRL) for 5 min. The retinal cells were dissociated by repeat pipetting in 1 mg/ml trypsin inhibitor (Sigma-Aldrich). Single cells were collected by passing through the 40 μ m cell strainer and cultured in the low adherent dish with Neurobasal A medium (Gibco BRL) supplemented with 1 × N2 (Gibco BRL), 1% bovine serum albumin (Sigma-Aldrich), 20 nM progesterone (Sigma-Aldrich), 20 ng/ml bFGF (PeproTech), 30 ng/ml epidermal growth factor (EGF; PeproTech) and 2 μ g/ml heparin (Sigma-Aldrich) for 3 days. Neurospheres were collected, transfected with miR-132 expression plasmid (OriGene, Rockville, MD) and cultured in the differentiation medium (Neurobasal A medium with 1x B27) for 7 days. RNA of the differentiated cells was collected and isolated as previously mentioned. Expression analysis on Pax6, Brn3a, Brn3b, Rho and Gfap genes was performed by Sybr green PCR with specific primers (Supplementary Table 1). The expression level was normalized by the housekeeping gene (Gapdh) and compared to that of the empty vector (pCMVMIR)-transfected cells. All rats were treated according to the guidelines of the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The experimental protocol was approved by the Animal Experimentation Ethics Committee of the Chinese University of Hong Kong. microRNA-132 transfection analysis. The miR-132 expression plasmid based on the PCMVMIR cloning vector was purchased from the commercially available source (OriGene). The miR-132 expression plasmid was transfected into human PDLSCs through the TransIT ® -LT1 transfection reagent (Mirus Bio LLC, Madison, WI) according to the manufacturer's protocol. Five days after transfection, protein from the transfected cell was collected for CUX1 expression analysis by immunoblotting as previously mentioned. β -actin was used as housekeeping protein for normalization. The expression level of the miR-132-transfected cells was compared to that of pCMVmiR-transfected cells.