SOX2 Is Required for Adult Human Müller Stem Cell Survival and Maintenance of Progenicity In Vitro

Abstract

Purpose.: SOX2, a high-mobility group transcription factor, is expressed by retinal progenitors during development. It has been associated with the ability of progenitor cells to differentiate into retinal neurons and is highly expressed by human Müller stem cells (hMSCs) in culture. The authors investigated the role of this factor in the maintenance of progenicity and neural differentiation of hMSCs in vitro.

Methods.: SOX2 silencing was induced by transfection of hMSCs in culture with two pGSU6-GFP SOX2 silencing constructs and a scrambled control vector. Silencing was confirmed by examination of gene and protein expression coding for SOX2. Effects of SOX2 downregulation were investigated by expression of proliferation (Ki67) and apoptotic (TUNEL, caspase) cell markers and by the expression of markers of retinal neurons (HuD, βIII tubulin, rhodopsin, BRN3B, ISL1), glia (vimentin), and the progenitor marker PAX6.

Results.: SOX2 silencing caused hMSCs to rapidly adopt a neural-like morphology and was accompanied by the upregulation of specific markers of retinal neurons, including βIII tubulin, rhodopsin, BRN3B, and ISL1, and by the downregulation of the neural progenitor marker PAX6 and the glial cell marker vimentin. Interestingly, SOX2 silencing induced apoptosis, suggesting a crucial role of this factor on hMSC survival in vitro.

Conclusions.: These in vitro results parallel that seen when Sox2 is silenced in neural stem cells of lower species during development, and they suggest that Sox2 may have an important role in adult hMSC differentiation into retinal neurons in vitro.

During development, Müller glia and retinal neurons share a common progenitor that is multipotent at all stages of retinal histogenesis. ¹ Müller cells constitute a source of neural retinal progenitors in the adult zebrafish ² and the postnatal and damaged adult chick ³ , rat ^4,5 and mouse retinas. ^6,7 Recent studies have demonstrated that the adult human retina also harbors a population of Müller glia with stem cell characteristics. ^8,9 Human Müller stem cells (hMSCs) express various markers of neural stem cells in vitro and in situ, including the transcription factor SOX2. ^8,9 This gene encodes a member of the SRY-related high-mobility group (HMG) family of transcription factors involved in regulation of embryonic development and determination of cell fate. ¹⁰ SOX2 plays a role in many developmental processes and is essential for the development of the CNS, ¹¹ sensory taste bud cells, ¹² sensory cells of the inner ear and eye, ¹³ and branching morphogenesis and epithelial cell differentiation in the lung. ¹⁴ Sox2 is also expressed in the adult brain where neural stem cells divide at a low rate, both renewing themselves and generating neural precursors. ¹⁵ In culture, SOX2-positive hMSCs show functional properties of neural stem cells, including self-renewal and multipotency. ⁸ Sox2 has also been shown, in combination with the expression of the genes Oct 3/4, cMyc, and Klf to induce pluripotent stem cells from various somatic cells, including fibroblasts and epithelial cells. ¹⁶  

Inhibition of SOX2 by injection of antisense morpholinos or dominant-negative forms of SOX2 into Xenopus embryos results in reduced or missing eyes. ¹⁷ Within the retina Sox2 is expressed in proliferating cells and is downregulated by all cell types as they start to migrate to the different retinal areas. ¹⁸ Sox2 is also required for the development of amacrine cells and Müller glia, ¹⁹ and the inhibition of Sox2 signaling results in neural progenitor cells delaminating from the ventricular zone and exiting the cell cycle prematurely. This is associated with general loss of pan-neural and regional progenitor markers and the onset of expression of early neuronal differentiation markers. ^15,16 In the adult Sox2 mutant mouse, precursor cell proliferation and generation of new neurons in neurogenic areas are decreased, whereas GFAP/Nestin-positive hippocampal cells are greatly diminished. ²⁰ Given that hMSCs can be made to differentiate in vitro into cells that exhibit characteristics of retinal neurons ⁸ and given the importance of this gene during neural retinal development and its well-recognized function in the maintenance of “stemness” on neural stem cells, we investigated whether SOX2 may also play a role in the maintenance of progenicity of adult hMSCs in culture and whether downregulation of this gene may modify the neural differentiation ability of these cells in vitro. 

Methods

Preparation of Silencing Constructs

SOX2 was silenced using short hairpin RNA (shRNA) constructs cloned into the p6GSU6-GFP vector (Genesilencer shRNA System; Gene Therapy Systems Inc., San Diego, CA). Two regions of human SOX2 were targeted that are part of the sequences coding for different protein domains using the GenBank database (Table 1). Oligonucleotide sequences were purchased from Invitrogen (Carlsbad, CA) and resuspended to a concentration of 1 μg/mL. Silencing and control constructs were prepared according to the manufacturer's instructions as follows: Equal amounts of forward and reverse oligonucleotides were mixed with annealing buffer (supplied in the vector kit), heated to 90°C, and allowed to cool slowly at room temperature overnight. PCR grade water was added to achieve a final oligo concentration of 10 ng/mL. Annealed oligonucleotides were ligated into the linearized vector using T4 DNA Ligase (catalog no. 15224–041; Invitrogen). Ligation was confirmed by restriction enzyme digest with HindIII (catalog no. R0104L; New England Biolabs, Beverly, MA) in the center of the inserted sequence, which yielded a product of 5165 base pairs corresponding to the linearized vector. In addition, the presence of the shRNA sequences was also confirmed by sequencing of plasmid DNA using the pGSU6-GFP sequencing primer (AATGAGGAAATTGCATCGCATTGTCTGAGTAG) in accordance with the manufacturer's handbook and was sequenced on a DNA analyzer (3730 DNA Analyzer; Applied Biosystems, Foster City, CA). Sequences were visualized using the freeware Chromas Lite Browser (version 2.01; www.technelysium.com.au). 

Table 1.

 

SOX2 shRNA Construct Sequences

Construct	Sequence
shRNA1a	5′ ATCC GT GCC TTC ATG GTG TGG TCC GAAGCTTG GGA CCA CAC CAT GAAGGC ATTTTTGGAAG 3′
shRNA1b	5′ AATTCTTCCAAAAAAT GCC TTC ATG GTG TGGTCCCAAGCTTCGGA CCA CAC CAT GAA GGC ACG 3′
shRNA2a	5′ GATCC GC GGC AGC TAC AGC ATG ATG GAAGCTTG CAT CAT GCT GTA GCT GCC GTTTTTTGGAAG 3′
shRNA2b	5′ AATTCTTCCAAAAAAC GGC AGC TAC AGC ATG ATG CAAGCTTC CAT CAT GCT GTA GCT GCC GCG 3′
Control-a	5′ GATCC GC CGG TAC CGA GTA AGC GTA GAAGCTTG TAC GCT TAC TCG GTA CCG GTTTTTTGGAAG 3′
Control-b	5′ AATTCTTCCAAAAAAC CGG TAC CGA CTA AGC GTA CAAGCTTC TAC GCT TAC TCG GTA CCG GCG 3′

Müller Stem Cell Culture and Transfection with SOX2 shRNA Constructs

In this study human Müller stem cell lines isolated from the human retina (cell lines MIO-M1, MIO-M2, and MIO-M5) were cultured under normal and differentiating conditions, as previously described. ^8,21 Briefly, cells were grown in 75-cm² flasks up to approximately 70% confluence. They were trypsinized, pelleted, and resuspended to a concentration of 3 × 10⁶ cells/mL in hypo-osmotic electroporation buffer (catalog no. 940002001; Eppendorf, Hamburg, Germany). The silencing construct was added at a concentration of 20 μg/mL to 800 μL cell suspension, and the solution was transferred to an 800-μL, 4-mm gap electroporation cuvette (catalog no. 940001021; Eppendorf). Cells were electroporated using an electroporation system (Multiporator; (catalog no. 4308000015; Eppendorf) at 2 pulses (1 minute apart) of 520 V for 100 μs. Cuvettes were transferred to 37°C for 5 minutes and then transferred into 25-cm² flasks or chamber well slides (Permanox; Nunc, Leicestershire, UK). Media were changed 24 hours after transfection, and G418 antibiotic was added at a concentration of 50 μg/mL (catalog no. G8168; Sigma) to obtain enriched populations of SOX2-silenced cells. Cells were visualized for GFP fluorescence using an epifluorescence microscope (Leica, Wetzlar, Germany). On average, cells used for RT-PCR or Western blot analysis showed approximately 80% GFP positivity. 

To quantify the number of transfected cells displaying a neural morphology, 100 cells from three different experiments were analyzed using the cell counter plugin of the ImageJ software (developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html; version 3.9u). The number of cells with primary neurites (long thin extensions from the cell body) and secondary neurites (branching from the primary neurites) were counted. 

Immunostaining of Cells

Cells were fixed in 4% paraformaldehyde (PFA) for 10 minutes, cryopreserved with 30% sucrose, and kept at −20°C until use. After defrosting, slides were blocked for 1 hour (0.5% Blocking Solution; Roche, Welwyn Garden City, UK; http://www.roche-applied-science.com) and were immunostained using our published methods. ^8,9,22 The primary antibodies used were BRN3B βIII tubulin (05–661; mouse; 1:200; Chemicon, Temecula, CA), Ki67 (NCL-Ki67p; rabbit; 1:1000; Novacastra, Newcastle upon Tyne, UK), SOX2 (goat; 1:100; SC17319; Santa Cruz Biotechnology, Santa Cruz, CA), BRN3B (N15; SC-31,987; 1:200; goat; Santa Cruz Biotechnology), ISL1 (mouse; 1:100; Developmental Studies Hybridoma Bank, Iowa City, IA), rhodopsin (mouse; clone 4D2, kind gift of Robert Molday, University of British Columbia), SOX2 (goat; 1:1000; Santa Cruz Biotechnology), and vimentin (rabbit; 1:1000; Santa Cruz Biotechnology). Specific binding of primary antibodies was detected using donkey anti-IgG labeled with AlexaFluor 488 or 555 (Molecular Probes, Invitrogen) reacting with the species in which the primary antibody was raised for 1 hour at room temperature. DAPI (Sigma) was used to stain cell nuclei, and slides were mounted with mounting medium (Vectashield; Vector Laboratories, Burlingame, CA). Fluorescent images were recorded using a confocal microscope (TCS SP2 AOBS; Leica). 

Apoptosis Detection

For the tunnel assay, an in situ apoptosis kit (ApopTag; catalog no. S7160; Chemicon) was used to detect fragmented cell nuclei. Briefly, cells were fixed in 1% PFA for 10 minutes and were postfixed in a 2:1 solution of ethanol/acetic acid. They were then incubated for 1 hour with TdT enzyme at 37°C. After the enzymatic reaction was stopped, cells were incubated with anti–digoxigenin antibody (catalog no. ab420; Abcam, Cambridge, UK) at a 1:1000 dilution for 30 minutes, followed by a 1-hour incubation with a secondary anti-mouse antibody labeled with Alexa555 (as described). Cells were costained with DAPI (as described) and mounted (Vectashield; Vector Laboratories). 

We used a detection kit (image-iT LIVE Red Poly-Caspase; catalog no. I35102; Invitrogen) to detect cells in the early stages of apoptosis without denaturing the GFP protein expressed by transfected cells. This kit uses a fluorescence-labeled probe that binds to the reactive center of activated caspases (FLICA). ²³ The FLICA inhibitor consists of a caspase-selective amino acid sequence (valine-alanine-aspartic acid) that reacts with most active caspases and a red fluorescent sulforhodamine reporter group. Cells were incubated with this reagent for 1 hour at 37°C. After this a Hoechst stain was used to label the nuclei of all cells. Slides were mounted using the wash buffer supplied in the kit and visualized immediately by confocal microscopy (as described). 

RT-PCR

Total RNA was extracted with a kit (RNeasy; Qiagen, Valencia, CA; http://www.qiagen.com) according to the manufacturer's instructions. Briefly, RNA (1 μg) was transcribed into cDNA using reaction buffer (AMV Reverse Transcriptase; Roche) and our standardized methods. ²² The reaction was performed in a final volume of 20 μL consisting of 5 mM MgCl₂, 1 mM dNTP, 1 U/μL RNase inhibitor, 0.8 U/μL reaction buffer (AMV Reverse Transcriptase; Roche), and 80 ng/μL oligo dT-15 primers. The mixture was incubated for 10 minutes at 25°C, 60 minutes at 42°C, 5 minutes at 99°C, and 5 minutes at 4°C in a thermal cycler (Mastercycler; Eppendorf; http://www.Eppendorf.co.uk). cDNA (5 μL) was used for PCR reactions using a PCR kit (High Fidelity; Roche). Amplification was performed in a final reaction volume of 25 μL consisting of 1.5 mM MgCl₂, 0.2 mM dNTP, 2.5 U Taq DNA polymerase (Expand HiFi; Roche), 0.4 μM primers in 50 mM KCl (Table 2 for primer sequences), and 10 mM Tris/HCl, pH 8.0. Primer sequences for the various neuronal and progenitor markers investigated are listed in Table 2. The mixture was incubated at 94°C for 2 minutes followed by 30 cycles under the following conditions: 94°C for 30 seconds, 60°C for 30 seconds, 72°C for 1 minute, and 1 cycle of 72°C for 5 minutes. An annealing temperature of 60°C was used for all primers. Products were run on 1% agarose gel containing a 1 in 15,000 dilution of nucleic acid stain (SYBR Gold; Invitrogen). GAPDH primers were used as a loading control. 

Table 2.

 

Primer Sequences

Name	Forward Sequence	Reverse Sequence	cDNA Location (base pairs)	Product Size (base pairs)
GAPDH	CCACCCATGGCAAATTCCATGGCA	TCTAGACGGCAGGTCAGGTCCACC	188–785	598
Pax6	AGATGAGGCTCAAATGCGAC	GTTGGTAGACACTGGTGCTG	1102–1385	302
Rhodopsin	GCTTCCCCATCAACTTCCTCA	AGTATCCATGCAGAGAGGTGTAG	152–285	156
Sox2	GGCAGCTACAGCATGATGC	TCGGACTTGACCACCGAAC	932–1149	236
HuD	GAAACTGTCCTTCTCCCATGC	GATTGAGGCAGAGCTCGGAC	310–611	301

Western Blot Analysis

Cell lysates were prepared using RIPA buffer containing a protease inhibitor cocktail (Sigma). Gels and buffer systems (NuPAGE; Invitrogen) were used for Western blot analysis, as previously described. ²² Protein concentration was estimated using the Bradford assay system (Bio-Rad, Hemel Hempstead, UK; www.biorad.com). Briefly, protein concentrations loaded onto the electrophoresis gel were adjusted according to the relative absorbance readings. In addition, a loading control (β-actin) was also used in each experiment. Thirty microliters of a loading sample was prepared with 3 μL reducing agent (10×), 7.5 μL loading buffer (LDS 4×), and a maximum of 19.5 μL protein sample. Proteins were denatured at 80°C for 10 minutes before loading onto 10% bis-tris gels with MOPS buffer containing antioxidant. Gels were run at 180 V for 50 minutes, after which they were transferred onto polyvinylidene difluoride membranes (Hybond-P; Amersham, Little Chalfont, UK; http://www.gelifesciences.com) at 35 V for 90 minutes. Membranes were blocked in TBS + 0.1% Tween-20 with 5% skimmed milk and 3% fetal bovine serum at 37°C for 1 hour. Primary antibodies used were BRN3B (N15; SC-31,987; 1:200; goat; Santa Cruz Biotechnology), ISL1 (mouse; 1:100; Developmental Studies Hybridoma Bank), S-opsin (N-20; goat; 1:200; Santa Cruz Biotechnology), SOX2 (goat; 1:1000; Santa Cruz Biotechnology), vimentin (rabbit; 1:1000; Santa Cruz Biotechnology), and β-actin (1:5000; monoclonal; Sigma) diluted in blocking buffer, after which the membrane was incubated at 4°C overnight before incubation with a secondary antibody conjugated to horseradish peroxidase (1:10,000; Jackson Laboratories, Bar Harbor, ME; http://www.jacksonimmuno.com/) for 1 hour at room temperature. The blot was visualized with enhanced chemiluminescence advanced detection reagent (GE Healthcare, Little Chalfont, UK; http://www.gelifesciences.com) using a Fujifilm imager (LAS-100; Fujifilm, Tokyo, Japan; www.fujifilm.com). 

Statistical Analysis

The integrated optical density of each band was calculated using ImageJ software (version 3.9u). Histograms were generated showing the pixel intensities of each band. The optical density of each band was normalized by dividing the optical density of the sample by the optical density of its corresponding control gene (GAPDH or β-actin) band. Statistical analysis of all results was carried out with statistical software (Prism 4; GraphPad, San Diego, CA). At least three different experiments were used for statistical analysis. SEM was plotted as error bars on bar charts, and an independent t-test was used to test the significance of the results. P < 0.05 was considered significant. 

Results

SOX2 Silencing in Human Müller Stem Cells

Transfection of hMSC lines with two SOX2 shRNA constructs caused downregulation in the expression of this factor as judged by immunostaining for SOX2 3 days after transfection (Fig. 1A) and RT-PCR and Western blot analysis from RNA and protein extracted from cells selected by G418 treatment 1 week after transfection (Figs. 1B, 1C). The shRNA2 construct was shown to be more effective at silencing SOX2 than shRNA1. Semiquantitative RT-PCR showed that shRNA1 caused a decrease in SOX2 mRNA levels by approximately 40%, whereas shRNA2 induced a decrease of 60% in mRNA levels after 5 days after transfection (Fig. 1B). Western blot analysis showed a marked reduction (approximately 60%) in SOX2 protein expression in cells transfected with both shRNA constructs when compared with the control construct (Fig. 1C). However, no statistically significant difference in the levels of SOX2 protein expression was observed between cells transfected with the two silencing constructs (Fig. 1D). 

Figure 1.

 

SOX2 silencing in human Müller stem cells. (A) Immunostaining for SOX2 showing a marked decrease in the expression of this factor by cells transfected with SOX2 shRNA 3 days after transfection. Decrease in SOX2 expression was not apparent after transfection with the scrambled control construct. (B) RT-PCR gel bands showing SOX2 mRNA expression by cells transfected with the two silencing and control constructs, selected for SOX2 shRNA expression with G418 for 1 week. Histogram illustrates the average band densities for three cell lines, showing a significant decrease in SOX2 gene expression by cells transfected with both constructs (*P = 0.0384, shRNA1 vs. scrambled; **P = 0.0047, shRNA2 vs. scrambled). shRNA2 was more efficient at silencing SOX2 than shRNA1 (***P = 0.003). (C) Representative Western blot for SOX2 protein and histogram of the average band densities from three cell lines cultured for 7 days with G418. SOX2 was significantly downregulated (by 60%) in cells transfected with both shRNA constructs compared with the scrambled control (*P = 0.0188, shRNA1 vs. scrambled; **P = 0.0226, shRNA2 vs. scrambled; n = 3. (D) Cells expressing SOX2 shRNA (GFP staining) adopted neural-like morphology 1 week after transfection, as shown by the acquisition of long processes reminiscent of neurites. They also developed bulblike structures resembling growth cones (white arrows). Cells transfected with the control scrambled construct did not acquire a characteristic neural-like morphology. Scale bars, 50 μm. (E) Quantification of the number of cells displaying a neural morphology at 7 days after transfection showed that the SOX2-silencing constructs induced cells to extend more primary and secondary neurites (where long thin processes were considered as neurites and secondary neurites were branches of primary neurites; ***P < 0.04, shRNA1 vs. scrambled and shRNA2 vs. scrambled). A significantly greater percentage of cells transfected with shRNA2 extended primary and secondary neurites compared with cells containing shRNA2 (*P = 0.04, shRNA1 vs. shRNA2-primary;** P = 0.009, shRNA1 vs. shRNA2-secondary; n = 3).

Cells transfected with SOX2 shRNA constructs and selected with G418 for 1 week showed a dramatic change in cell morphology, often extending multiple long processes, reminiscent of neurites (Fig. 1D). Some cells had also bulblike structures along the processes, resembling growth cones. These changes suggested that cells were acquiring a neuronal phenotype because of SOX2 downregulation. Most cells displaying these neuronal morphologies did not survive for more than 2 weeks in culture, and only a few cells displaying a neural phenotype could be seen after 2 weeks after transfection. Quantification of the number of cells displaying neural morphologies showed that a significantly greater number of cells transfected with silencing constructs extended primary and secondary processes (neurites) compared with control cells (Fig. 1E). In addition, a greater number of cells transfected with shRNA2 extended both primary and secondary neurites (50%) compared with cells expressing shRNA1 (15%; Fig. 1E). Cells transfected with the scrambled control construct did not exhibit this neural morphology after 1 week in culture (Fig. 1D), suggesting that the acquisition of a neural-like morphology by hMSCs was due to the silencing of SOX2 and not to the transfection itself. In addition, cells transfected with the scrambled control did survive and continued dividing indefinitely in culture (data not shown). 

Figure 2.

 

Effect of SOX2 silencing on the expression of progenitor, glial, and neuronal cell markers. (A) RT-PCR gel bands showing decreased expression of PAX6 mRNA by cells transfected with the two SOX2-silencing and control constructs. Histogram illustrates the average band densities for three cell lines, showing a significant decrease in PAX6 gene expression by cells transfected with both constructs (*P = 0.005, shRNA1 vs. scrambled; **P = 0.0006, shRNA2 vs. scrambled). (B) Western blot illustrating the expression of vimentin by Müller stem cells transfected with SOX2 shRNA and control constructs. Histogram of band densities normalized to β-actin shows that although both shRNA constructs decreased the expression of vimentin, only the effect of shRNA2 was significant (*P = 0.0052, shRNA2 vs. scrambled). In addition, the effect of shRNA2 was significantly greater than that of shRNA1 (**P = 0.0034, shRNA1 vs. shRNA2; n = 3). (C) Representative RT-PCR gel bands showing expression of HuD mRNA by cells transfected with the two SOX2-silencing and control constructs. Histogram illustrates the average band densities for three experiments, showing no changes in HuD gene expression by cells transfected with both shRNA constructs. (D) Confocal images of cells transfected with SOX2-silencing constructs (GFP, green) shows cells that acquired long processes reminiscent of neurons 7 days after transfection with shRNA2. They also expressed βIII tubulin (red). These features were not observed with the scrambled control construct. Scale bars, 20 μm.

Effect of SOX2 Silencing on Progenitor, Glial, and Neuronal Cell Marker Expression

If cells expressing silencing constructs were truly differentiating into retinal neurons, it would be expected that they would show downregulation of glial and progenitor cell markers while upregulating markers of mature neurons. On this basis, the effects of SOX2 shRNA on hMSCs was examined by investigating the expression of the progenitor marker PAX6, the glial marker vimentin, and the neuronal markers βIII tubulin and HuD to identify whether transfected cells were truly acquiring neural retinal characteristics.e Cells transfected with both SOX2-silencing constructs showed a significant decrease in the expression of PAX6 compared with cells transfected with the scrambled construct (Fig. 2A) at 7 days after transfection. ShRNA2 appeared to induce a greater downregulation of this gene than shRNA1, but this difference was not statistically significant (Fig. 2A). 

Vimentin, an intermediate filament protein, is only expressed by Müller glia in the neural retina. Interestingly, cells transfected with the SOX2-silencing shRNA2 showed a marked reduction in the expression of this protein, reaching approximately one-third of the level seen in control cells (Fig. 2B). Although a slight decrease in vimentin was observed in cells transfected with construct-1 after 7 days, this was not statistically significant (P = 0.052). Investigation of the expression of HuD, a marker of amacrine and ganglion cells, showed that SOX2 silencing did not modify the expression of mRNA coding for this molecule (Fig. 2C). Immunostaining for βIII tubulin showed that cells transfected with shRNA2 that developed neural-like morphology expressed this protein, which is characteristically expressed by developing neurons (Fig. 2D). These results suggest that as the transfected cells became neuronal in morphology, they gained expression of some neuronal markers while decreasing their expression of progenitor and glial markers. 

Figure 3.

 

Photoreceptor and ganglion cell marker expression by Müller stem cells transfected with SOX2 shRNA constructs. (A) Western blot analysis of cell lysates obtained 7 days after transfection and G418 selection show that silencing of SOX2 in Müller stem cells caused expression of rhodopsin, as identified by its characteristic multiple bands reactive with anti–rhodopsin antibody. Cells transfected with the scrambled control construct did not express this protein. Histogram of band densities normalized to β-actin shows that cells transfected with both constructs significantly increased the expression of this protein (*P = 0.0008, shRNA1 vs. scrambled; **P = 0.0001, shRNA2 vs. scrambled; n = 3). (B) Expression of BRN3B protein by SOX2 shRNA-transfected cells was significantly upregulated by both silencing constructs. Histogram of band densities normalized to β-actin showed significant upregulation of BRN3B protein expression (*P = 0.0021, shRNA1 vs. scrambled; **P = 0.0004, shRNA2 vs. scrambled; n = 3). (C) ISL1 protein expression by Müller stem cells transfected with SOX2 shRNA constructs. Representative Western blot for ISL1 showing that cells transfected with both silencing constructs increased expression of this protein. Cells transfected with the silencing construct did not express ISL1. Histogram of band densities normalized to β-actin shows that cells transfected with both constructs upregulate ISL1 protein expression (*P = 0.0007, shRNA1 vs. scrambled; **P = 0.0016, shRNA2 vs. scrambled; n = 3). (D) Confocal images of SOX2-silenced cells using shRNA2 showed that cells transfected with SOX2 shRNA (green) that acquired an elongated morphology expressed cytoplasmic rhodopsin as well as nuclear BRN3B and ISL1 (red).

Downregulation of SOX2 Increases Expression of Markers of Photoreceptor and Ganglion Cells

After 7 days' culture of hMSC transfected with SOX2 shRNA, cells were investigated for their expression of photoreceptor and ganglion cell markers. Western blot analysis showed that silencing of SOX2 caused an increase in the expression of the rod photoreceptor marker rhodopsin. After 7 days' culture, rhodopsin protein in cells transfected with shRNA1 and shRNA2 were significantly increased when compared with cells transfected with the scrambled control (Fig. 3A). This was illustrated by the appearance of multiple rhodopsin bands in the Western blots, which are characteristic of this protein. ²⁴ Levels of rhodopsin were upregulated by 50% in cells transfected with shRNA2 compared with cells transfected with the control construct (Fig. 3A; P = 0.0147). 

Cells transfected with both silencing constructs also showed an upregulation of BRN3B protein (Fig. 3B), a transcription factor expressed by differentiating retinal ganglion cells. ²⁵ Calculation of Western blot band densities revealed that the expression of this protein increased by a factor of 6 in cells transfected with SOX2 shRNA compared with cells expressing the control construct at 7 days after transfection (Fig. 3B). Although the difference was not statistically significant, shRNA2 appeared to induce slightly higher levels of BRN3B expression than shRNA1. The expression of Islet1 (ISL1), another protein marker expressed by differentiating retinal ganglion cells ²⁶ and amacrine cells, ¹⁹ was also investigated. Levels of ISL1 showed a fourfold increase in cells expressing SOX2 shRNA compared with control cells (Fig. 3C). As seen with BRN3B expression, cells transfected with shRNA2 expressed slightly more ISL1 than those transfected with shRNA1. These results indicate that silencing of SOX2 induced Müller stem cells to express characteristic markers of retinal neurons. 

Immunostaining of cells transfected with shRNA showed rhodopsin expression in the cell bodies of some cells exhibiting a neural-like morphology (Fig. 3D). This is in agreement with observations that although rod outer segments are not formed in photoreceptors cultured in vitro, rhodopsin protein localizes to the cell body and processes of cells in culture. ²⁷ Cells transfected with the SOX2-silencing constructs that extended long processes were also positive for nuclear BRN3B and ISL1 (Fig. 3D). This pattern of nuclear staining for these two molecules is characteristic of retinal ganglion cells. ^25,26  

Effect of SOX2 Downregulation on hMSC Proliferation and Survival

Previous studies have shown that cells silenced for SOX2 exit the cell cycle and eventually undergo apoptosis. ^13,20 To determine whether SOX2 shRNA was causing Müller stem cells to cease proliferating, cells were investigated for the expression of Ki67. Ki67 is present in cells in active phases of the cell cycle (G1, S, G2, and mitosis) and localizes to the nucleus. Quantification of SOX2 shRNA-transfected cells stained for Ki67 revealed that 12% to 15% of cells transfected with the control construct were within the active phases of the cell cycle 1 week after transfection (Figs. 4A, 4B). In contrast, only 3% of cells transfected with shRNA1 and 1% of cells transfected with shRNA2 were actively dividing (Figs. 4A, 4B). These results suggest that the downregulation of SOX2 induces adult human Müller stem cells to exit the cell cycle. 

Figure 4.

 

Effect of SOX2 downregulation on Müller stem cell proliferation. (A) Confocal images of hMSCs transfected with the SOX2 shRNA1 and control construct (green) stained for the proliferating antigen Ki67 (red). Very few cells transfected with the silencing constructs 1 and 2 stained for this molecule (white arrows), whereas a larger proportion of cells transfected with the control construct expressed Ki67 (B) Histogram represents the percentage of transfected cells staining positive for Ki67 in populations of cells after 7 days' transfection and G418 selection. Only 1% of cells were Ki67-positive when transfected with shRNA2, and 4% showed staining when transfected with shRNA1 (*P = 0.0033, shRNA1 vs. scrambled; **P = 0.0006, shRNA2 vs. scrambled; n = 6). Scale bars, 50 μm.

Along with a reduction in proliferation, cells transfected with SOX2 shRNA often had fragmented nuclei (revealed by DAPI staining), suggesting that these cells were becoming apoptotic (Fig. 5A, upper panel). On this basis, two different apoptosis assays were used to confirm these observations. A TUNEL assay was used to identify cells in the last stages of apoptosis (detecting strand breaks within the DNA), and a poly-caspase assay was used to identify those cells that were in the early stages of apoptosis. ²⁸ Unfortunately, a methanol/acetic acid fixation step in the protocol for this assay caused denaturation of the GFP protein expressed by transfected cells and, therefore, could not be visualized in TUNEL-stained cells. To ensure that nontransfected cells were not included in the analysis, G418 antibiotic was added to the cell cultures to select only transfected GFP-positive cells before the TUNEL assay. The TUNEL assay revealed that very few cells (approximately 1%) transfected with the control construct were apoptotic. In contrast, more than 20% of cells transfected with shRNA2 showed apoptosis (Fig. 5A). The increase in apoptosis observed in cells transfected with construct-2 was significantly increased when compared with cells transfected with shRNA1; only 6% to 8% of cells were positive for TUNEL staining (Fig. 5A). 

Figure 5.

 

Effect of SOX2 downregulation on hMSC death. (A, top) Fluorescent images of cells transfected with the GFP-shRNA2 (left) containing fragmented nuclei (white arrows) as shown by counterstaining with DAPI (right). Bottom: confocal images of cells transfected with GFP-shRNA2 and scrambled construct staining for TUNEL (red) at 1 week after transfection. Because GFP staining was reduced by acid treatment used in the TUNEL protocol, transfected cells were selected with G418 before the assay. Histogram shows a significant increase in the number of TUNEL-positive cells within the cell population transfected with the shRNA1 and shRNA2 constructs. Only a very small number of cells transfected with the control construct showed TUNEL staining (*P < 0.0001, shRNA1 vs. Scrambled, shRNA2 vs. Scrambled; **P = 0.0006, shRNA1 vs. shRNA2; n = 6). Scale bars, 50 μm. (B) Confocal images of SOX2 shRNA-transfected cells stained for poly-caspases (red). Cells transfected with the control construct (green) did not stain for caspases (top). In contrast, a large proportion of cells transfected with the silencing GFP-shRNA2 stained for these molecules (bottom). Histogram shows quantification of caspase-positive cells after transfection with the different constructs (*P < 0.0001, shRNA1 vs. scrambled and shRNA2 vs. scrambled; **P = 0.0003, shRNA1 vs. shRNA2; n = 6). Scale bars, 50 μm.

Using the poly-caspase assay, we observed that between 30% and 40% of cells transfected with the silencing shRNA2 were in the early stages of apoptosis at 1 week after transfection (Fig. 5B). Levels of apoptotic cells in hMSC preparations transfected with construct-2 were significantly higher when compared with cells transfected with shRNA1; 10% of cells were apoptotic (P = 0.0003). In contrast, only 1% of cells transfected with the control construct showed caspase staining. These observations strongly indicate that silencing of SOX2 in human Müller stem cells induces a reduction in proliferation and an increase in cell death. 

Discussion

Although much research into the role of SOX2 has been performed in neural stem cells from small mammals and other species during retinal development, very little is known on the role of this transcription factor in adult neural stem cells from the human eye. The complex nature of SOX2 regulation has been shown by previous work that demonstrate the interactions of Sox2 with Pax6, ²⁹ Notch1, ¹⁷ and the Wnt pathway. ³⁰ In addition, the activity of this factor may be regulated through nuclear import and export ³¹ because it has been shown that translocation from the nucleus to the cytoplasm occurs at different stages of embryonic development. ¹³ Our previous studies have shown that a subpopulation of hMSCs from the adult eye exhibit properties of neural stem cells. ^8,9 They express SOX2 in both their nuclei and cytoplasm, suggesting that this factor may play an active role not only in the survival and progenicity of these cells but also in their ability to differentiate into retinal neurons in vitro. 

To determine the importance and role of SOX2 expression in hMSC maintenance and survival, this gene was silenced in these cells in vitro using shRNA. The technique proved to be a reliable method to downregulate the gene, with two silencing constructs producing various levels of decrease in SOX2 expression by hMSCs. Both constructs induced hMSCs to change their morphology in vitro and to adopt a neural-like morphology that was evident in some cells at 24 hours after transfection. We previously showed that FGF2 in the presence of basement membrane protein induces hMSCs to acquire a neural morphology. ⁸ However, the neural-like phenotype observed after transfection with SOX2 shRNA was far more dramatic than that seen when cells were cultured in the presence of FGF2. This was supported by observations that in addition to inducing changes in their morphology, silencing of SOX2 in Müller stem cells also resulted in the downregulation of Müller and neural progenitor markers and was accompanied by the upregulation of markers of mature retinal neurons, including rhodopsin, ISL1, and BRN3b. 

Cells transfected with the two SOX2-silencing constructs showed a decreased expression of the transcription factor PAX6, expressed by all retinal stem/progenitor cells. Pax6 has been shown to interact with the Sox2 N3 enhancer region during embryonic development of the visual system. ²⁹ Pax6 and Sox2 also interact with each other to induce the expression of genes needed for lens development. ³¹ Recently, Sox2 has shown to interact with the Pax6 promoter to drive luciferase expression in Y79 cells and to be required for the development of amacrine cells and Müller glia. ¹⁹ The observation that SOX2 silencing causes the downregulation of PAX6 mRNA but not mRNA coding for the amacrine cell marker HuD suggests that SOX2 downregulation may not induce the differentiation of adult Müller stem cells into amacrine cells or that different signaling pathways may be involved in the differentiation of this neural cell type. In this context it is important to highlight the fact that amacrine cells share the expression of various markers with ganglion, horizontal, and bipolar cells. ^{32 –34} Therefore, it is difficult to infer from this observation on the real role of SOX2 on the differentiation of adult Müller stem cells into amacrine cells. However, this is the first demonstration that SOX2 has been implicated in the downregulation of PAX6 in adult stem cells, and it is possible that SOX2 downregulation may also result in the downregulation of other progenitor genes expressed by hMSCs. 

Silencing of SOX2 in hMSCs induced the upregulation of protein coding for photoreceptor and ganglion cell markers by these cells. Although these cells express mRNA coding for mature and developing rods and cones genes in the absence of ECM and growth factors, they do not appear to express these proteins. Müller stem cells transfected with the two silencing constructs expressed rhodopsin within 7 days after transfection. In addition to expressing this photoreceptor marker, hMSCs transfected with SOX2 shRNA also expressed markers of postmitotic retinal ganglion cells. Cells transfected with silencing constructs showed not only the upregulation of the retinal ganglion cell marker BRN3B and ISL1 proteins (also expressed by developing amacrine and bipolar cells, ^25,26 but also the mRNA coding for these markers (data not shown). These results suggest that on downregulation of SOX2, Müller stem cells are able to differentiate toward different retinal neuronal fates. Adoption of a neural fate was further confirmed by the downregulation of the glial protein vimentin, expressed by virtually all Müller stem cells. In addition, the proliferation of cells expressing SOX2 shRNA, as determined by Ki67 reactivity, was greatly reduced when compared with control cells. Under control conditions approximately 30% of cells are Ki67 positive after 1 week in culture (data not shown). The decrease in the number of proliferating cells may be attributed to the processes of electroporation and transfection, both of which have been shown to decrease the number of cells in S-phase of the cell cycle, and it has been shown that it may take between 2 and 6 weeks for cells to return to their previous levels of division. ^35,36  

Previous studies have shown that neural stem cells from Sox2 knockdown mice are unable to differentiate into mature neurons and undergo apoptosis, whereas homozygous Sox2 mutant embryos do not develop. ³⁷ Moreover, studies in the chick embryo have provided evidence that constitutive expression of SOX2 inhibits neuronal differentiation, resulting in the maintenance of progenitor characteristics. ³⁸ Similar effects caused by SOX2 downregulation were observed in this study, in which TUNEL and caspase assays revealed that hMSCs transfected with the silencing constructs showed an increase in apoptosis. These observations suggest that the expression of SOX2 appears to be essential for human Müller stem cell survival in vitro and that downregulation of this gene is important for neural differentiation of adult hMSCs. Based on the present results, it is not possible to determine the cascade of events that lead to neural differentiation of hMSCs and suggest that exogenous factors may be required to maintain the viability of terminally differentiated neurons in vitro. It might be possible that partial downregulation of SOX2 may be required only to induce neural differentiation while promoting cell survival; further studies should address these issues. The present observations, however, provide insight into the activity of SOX2 in hMSCs and indicate the importance of this transcription factor in the regulation of adult Müller stem cell survival and progenicity. 

Footnotes

 Supported by the Helen Hamlyn Trust in memory of Paul Hamlyn; Medical Research Council Grants 90465 and G0900002; The Henry Smith Charity; Fight for Sight (through a donation from Tony Bickford); the National Institute for Health Research Biomedical Research Centre at Moorfields Eye Hospital; the UCL Institute of Ophthalmology; and the Inlaks Foundation, India (SS).

Footnotes

 Disclosure: B. Bhatia, None; S. Singhal, None; D.N. Tadman, None; P.T. Khaw, None; G.A. Limb, None