Derivation of Human Differential Photoreceptor-like Cells from the Iris by Defined Combinations of CRX, RX and NEUROD

Examples of direct differentiation by defined transcription factors have been provided for beta-cells, cardiomyocytes and neurons. In the human visual system, there are four kinds of photoreceptors in the retina. Neural retina and iris-pigmented epithelium (IPE) share a common developmental origin, leading us to test whether human iris cells could differentiate to retinal neurons. We here define the transcription factor combinations that can determine human photoreceptor cell fate. Expression of rhodopsin, blue opsin and green/red opsin in induced photoreceptor cells were dependent on combinations of transcription factors: A combination of CRX and NEUROD induced rhodopsin and blue opsin, but did not induce green opsin; a combination of CRX and RX induced blue opsin and green/red opsin, but did not induce rhodopsin. Phototransduction-related genes as well as opsin genes were up-regulated in those cells. Functional analysis; i.e. patch clamp recordings, clearly revealed that generated photoreceptor cells, induced by CRX, RX and NEUROD, responded to light. The response was an inward current instead of the typical outward current. These data suggest that photosensitive photoreceptor cells can be generated by combinations of transcription factors. The combination of CRX and RX generate immature photoreceptors: and additional NEUROD promotes maturation. These findings contribute substantially to a major advance toward eventual cell-based therapy for retinal degenerative diseases.


Introduction
The possibility of redirecting cell differentiation by overexpression of genes was suggested by Weintraub with the identification of the ''master gene,'' MyoD [1]. The process was thought to involve reversion to a less differentiated state, a kind of de-differentiation, before the new cell type is formed. Another process has since been introduced, the concept of ''direct conversion'' or ''direct reprogramming'' without de-differentiation. This process is thought to be direct lineage switching [2] rather than lineage switching back to a branch point and out again in a different direction. ''Direct conversion'' has been shown in beta-cells, cardiomyocytes and neurons: A specific combination of three transcription factors (Ngn3, Pdx1 and MafA) reprogram differentiated pancreatic exocrine cells in adult mice into cells that closely resemble beta cells [3]; a combination of three factors (Gata4, Tbx5 and Baf60c) induces non-cardiac mesoderm to differentiate directly into contractile cardiomyocytes [4]; and a combination of three factors (Ascl1, Brn2 and Myt1l) converts mouse fibroblasts into functional neurons [5]. In this study, we employed the strategy of ''direct reprogramming'' to generate retinal photoreceptor cells from human somatic cells.
Several retinal diseases, including retinitis pigmentosa, agerelated macular degeneration and cone dystrophy, lead to loss of vision due to loss of photoreceptors and retinal pigment epithelium (RPE). Gene therapy has been implicated for Leber's congenital amaurosis [6]. Another promising therapeutic strategy is to transplant functional photoreceptor cells and retinal pigment epithelial cells. Sheets of human fetal neural retina with retinal pigment epithelium [7] and ES cell-derived photoreceptors [8] have been implicated for use as sources for photoreceptor cells. And human ES cell-derived RPE has recently been implicated to patients with macular degeneration [9]. However, the use of human embryos faces ethical controversies that prevent the widespread applications of human fetal tissues and human ES cells. A way to circumvent these issues is to induce photoreceptorspecific phenotypes by direct reprogramming of somatic cells of the patients. During vertebrate eye development, the inner layer of the optic cup differentiates into the neural retina and irispigmented epithelium (IPE). This common developmental origin led us to test whether iris cells could transdifferentiate to retinal neurons and thus be a candidate source of cells for transplantation. We here define the combinations of transcription factors that induce light responsive photoreceptor-like cells in humans.

Cultivation of iris-derived cells
The iris pieces were cut into smaller pieces and served as explants culture. Cells derived from the iris pieces are designated as ''iris cells''. Iris pigment epithelial cells (IPE cells) were isolated from iris tissues using dispase and trypsin. Residual iris pieces after removal of IPE were cut into smaller pieces and served as explant culture. Outgrowing cells from the explant cultures were designated as ''iris-stromal (IS) cells'' (Fig. 1A). Ciliary epithelial cells were isolated from pars plana and pars plicata in the same manner as IPE cells. We then performed Southern blot analysis and nucleotide sequencing to investigate whether the RB gene was deleted or mutated, because some irreversibly de-identified irisderived cells were from the patients with retinoblastoma. Southern blot analysis revealed that the RB gene was not deleted or rearranged in any of the iris-derived cells examined (Fig. S1). Sequencing analysis revealed that cDNAs of the RB gene did not have deletions or mutations at the nucleotide level.

Cultured iris cells show phenotypes of retinal glia and progenitor
Iris cells were immunocytochemically positive for glial cell-and neural stem cell-markers (Fig. 1A, 1B). RT-PCR analysis revealed that these cells expressed markers for glial cells and neural stem cells, indicating that the iris has common features with the retinal glia (Fig. 1C). After neural induction with the B27 medium, rhodopsin was not induced (Fig. 1C). After retinal induction with the R1 medium, green/red opsin was up-regulated significantly but blue opsin and rhodopsin were not up-regulated (Fig. 1D).

Iris cells are induced into a rod-or cone-specific phenotype by defined transcription factors
We selected six genes, SIX3, PAX6, RX, CRX, NRL, and NEUROD, as candidate factors that may contribute to induce photoreceptor-specific phenotypes in iris cells, on the basis that such factors play pivotal roles in the development of photoreceptors. Iris cells were transfected with these genes and were examined for inducible expression of photoreceptor-specific genes in those cells. Transduction of a single gene for SIX3, PAX6, RX, CRX, NRL, or NEUROD induced neither rod-nor cone-specific phenotypes in iris cells, but the six genes together up-regulated blue opsin and rhodopsin (Fig. S2). To determine which of the six candidates were critical, we tested the effect of withdrawal of individual factors from the pool of transduced candidate genes on expression of the opsin genes. We identified two genes, NEUROD and CRX, which were essential for photoreceptor induction; individual withdrawal of NEUROD resulted in loss of expression of rhodopsin and withdrawal of CRX resulted in loss of blue opsin.
Then, we tested the combination of only two genes, CRX and NEUROD ( Fig. 2A, 2B). The combination of CRX and NEUROD induced rod photoreceptor specific genes including rhodopsin and other phototransduction genes. After transduction of CRX and NEUROD, immunostaining showed that 38% of total cells were rhodopsin-positive cells (3,750 cells) (Fig. 2C, Fig. S3). However, this combination did not induce the red opsin gene. Addition of RX to the combination of CRX and NEUROD augmented blue opsin expression (Fig. 2B). After transduction with CRX, RX and NEUROD, rhodopsin-positive, blue opsin-positive and green/red opsin-positive cells were 29% (per 954 cells), 37% (per 235 cells) and 25% (per 193 cells) of total cells, respectively, by immunostaining. Hybrid photoreceptor cells were also detected by doublestaining immunocytochemistry (Fig. S4). We then investigated combinations of transcription factors that induce specific types of photoreceptor cells. A combinational approach showed that combination of CRX and RX was sufficient to induce green/red opsin and other cone-specific genes (Fig. 2D, Fig. S2). PAX6(+5a) did not influence cone-related gene induction (Fig. 2E). Expression levels of rhodopsin and blue opsin reached a maximum level by one week after gene transduction and remained unchanged up to 3 weeks. Expression of green/red opsin reached a maximum level 3 days after gene transduction (Fig. 2F). Expression levels of opsinand phototransduction-related genes were quantitated (Fig. 2G). NEUROD significantly decreased expression of the cone-specific genes, i.e. genes for green opsin and cone channel B3 (CNGB3) in human iris cells (p,0.005). On the other hand, it was clearly demonstrated that expression of rhodopsin and S-antigen, which are specifically expressed in rod photoreceptors, were much higher in CRX, RX and NEUROD-infected cells than in CRX and RXinfected cells (rhodopsin, p,0.05; S-antigen, p,0.005, Welch's ttest). Ultrastructural analysis revealed a cilia-associated structure, i.e. centriole, surrounded by mitochondria (Fig. S5).

Inhibition of factors by small interfering RNA (siRNA)
We performed RT-PCR to investigate if the transgenes continued to be expressed in the generated retinal cells (Fig. 2H, Table S1). The exogenous factors (transgenes) were clearly detected in induced retinal cells. Interestingly, the corresponding endogenous genes intiated expression in the induced retinal cells, similar to what is found in iPS cells. We then suppressed the CRX and NEUROD genes by siRNA (Fig. S6) to investigate the involvement of the genes in photoreceptor differentiation. Expression of the photoreceptor-specific/associated genes (blue opsin, s-antigen and recoverin) decreased significantly in siCRX and siNEUROD-transfected cells, compared to cells treated with control siRNA, suggesting that CRX and NEUROD are necessary for photoreceptor conversion.

Derivation of photoreceptor-like cells from IPE and IS cells
To investigate photoreceptor cell differentiation from other cell types, we isolated IPE and IS cells from iris tissues. Both cell types began to express opsin genes after transduction of CRX, RX and NEUROD genes (Fig. 3A, 3B). To determine if IPE and IS cells originated from neural ectoderm and neural crest cells, we investigated expression of neural crest marker genes. IPE and IS expressed these neural crest markers at high levels (Fig. 3C). These findings indicate that IS cells derived from neural crest cells, as well as IPE cells, could differentiate into photoreceptor-like cells. We also isolated ciliary epithelial cells from pars plicata and pars plana (Fig. 1A, 3D). Ciliary epithelial cells from pars plicata expressed rhodopsin, blue opsin, and green/red opsin at a high level after transduction with three genes (CRX, RX and NEUROD) or all six genes together (Fig. 3E). Retina-derived Müller glial cells expressed opsin genes after transduction of all genes (Fig. 3F).

Induced photoreceptor-like cells are photoresponsive in vitro
Light stimulation was applied to CRX, RX and NEURODinfected human iris-derived cells because these infected cells showed the most photoreceptor-like phenotypes by RT-PCR and immunocytochemistry. Both blue and green light stimulation produced inward current (Fig. 4A, 4B). Inward current continued to flow after the offset of light stimulation but only four cells showed partial or complete recovery within 60 sec after the cessation of light stimuli (n = 9), presumably reflecting the limited expression or absence of inactivation machinery. Inward current to blue light stimulation was observed in three out of four cells and inward current to green light stimulation was observed in six out of six cells. Light stimulation to non-infected control cells (blue, n = 2; green, n = 2) did not produce any inward current. These results indicate that the combination of CRX, RX and NEUROD transforms human iris-derived cells into photoresponsive photoreceptor-like cells in vitro, although the typical outward current of photoreceptor cells could not be detected. Since the light-induced inward current seemed to be mediated by melanopsin-associated phototransduction, we investigated expression of melanopsin by  Expression of the CRX, NEUROD and RX genes in the iris cells and transgene-induced cells was analyzed by RT-PCR, using the exogenous and endogenous gene-specific primers (Table S1). Human retina served a control for the endogenous genes. Equal amounts of RNAs were examined by expression of the G3PDH gene. doi:10.1371/journal.pone.0035611.g002 RT-PCR and immunocytochemistry. CRX, RX and NEURODinfected iris-derived cells expressed melanopsin (Fig. 4C, 4D), suggesting a larger contribution of melanopsin-associated inward current.

Discussion
This is the first report that functional photosensitive photoreceptor-like cells can be induced from human somatic cells. The present study shows that rod-and cone-photoreceptor-specific phenotypes were induced by transduction of a combination of CRX, RX and NEUROD genes, and that those cells responded to light electrophysiologically. In the retina, rod-and conephotoreceptors convert light information to electrical signals that are relayed to the brain through several interneurons. In the present study, a combination of CRX, RX and NEUROD induced all of the opsin genes: blue opsin, green/red opsin and rhodopsin (Table S2). On the other hand, a combination of CRX and RX induced only cone-specific opsin, and additional transduction of NEUROD up-regulated rod-specific opsin and rod-specific phototransduction related genes.
Rod photoreceptor generation from iris cells required NEUROD in our study. NeuroD is a regulator of both rod photoreceptors [10,11] and cone photoreceptors [12] during mouse development. NeuroD overexpression increases amacrine cells and rod photoreceptors, reduces bipolar cells, and inhibits formation of Müller glia. It has been known since the early 1960s that there is a defined sequence in formation of retinal neurons, which is largely conserved across vertebrates: Cone photoreceptors are generated during early stages of development, and most rod photoreceptors are generated in the latter half of the period of retinogenesis [13]. Similarly, cone photoreceptors are generated at the early stages during ES cell differentiation and rod photoreceptors are generated at a later stage. The present study and these previous reports suggest that NeuroD may work downstream to regulate the development of rod-photoreceptors. NeuroD generally functions in a cell cycle-specific manner, and promotes cell cycle exit [11]. Rod formation may thus be mediated via cessation of cell cycle by NeuroD at the later stage.
It has been hypothesized that retinal stem cells can be found in the ciliary body [14], postnatal retina, and the iris [15]. Pure populations of IPE cells isolated from rat and chicken irises were shown to demonstrate ''stemness'' [16]. A portion of purely isolated IPE cells of rodents, especially nestin-positive IPE cells, differentiated into multiple neuronal cell types, pan-neural marker-expressing cell types and retina-specific cell types without genetic manipulation. On the other hand, it has been shown that retinal stem cells are not present in the human iris [17,18]. The present study demonstrates that human iris cells expressed stem cell markers such as nestin, N-cadherin, Sox2, Musashi-1 and Pax6. Expression of stem cell markers in iris cells may be attributed to the cell source, i.e. cells from infants. However, photoreceptor cell differentiation with exogenously added chemicals and growth factors was limited; that is, only green/red opsin was induced (Fig. 1). Other experimental evidence has also suggested the limitation in mammals without genetic manipulation. Progenitor cells from the mammalian iris, pars plana, and ciliary body do not show a convincing immunoreactivity for rhodopsin, phosducin, recoverin, PKC, or RPE65 [19], but are induced into photoreceptor progeny with retinal transcription factors [20,21]. We first indicate that human IS cells that originate from neural crest (Fig. 3C), as well as IPE cells, differentiate into photoreceptor-like cells. Derivation of photoreceptor-like cells can be attributed to transgene-dependent differentiation of retinal progenitors that exist in the iris.
Our data show that induced photoreceptor-like cells have rodand cone-signaling-pathways by RT-PCR and immunocytochemistry. In addition, expression of melanopsin was also detected in these cells. Photostimulation of the rod-or cone-pathway produces hyperpolarizing responses, while activation of the melanopsinpathway produces depolarizing responses [22,23,24,25,26]. Melanopsin is intrinsically expressed in iris cells of the human (Fig. 4C), mouse [27] and Xenopus [24]. Melanopsin signaling has recently been reported to exist in both the iris and retina in mammals [27]. However, photostimulation did not produce any response in nontransfected human iris cells, suggesting the absence of phototransduction machineries per se. The light-induced depolarizing responses in infected cells indicate that phototransduction machinery for melanopsin-pathway was induced in infected cells. This is different from the results of infected monkey and rodent iris cells, where photostimulation produced hyperpolarizing responses  [20]. Our data demonstrate that transduction of three transcriptional factors strongly induces expression of blue opsin, which suggests a potential to produce hyperpolarizing responses. One plausible reason for the differences is that numbers of expressed phototransduction machineries for rod-or cone-pathways are not enough in those cells, e.g. outer segments were not observed at an ultrastructural level. Although the reason for depolarization in transfected cells cannot be fully explained so far, it is likely that melanopsin-associated photoresponses may overcome cone-or rod-mediated photoresponses.
In addition to revealing insights into retinal transdifferentiation, this study describes the development of a novel experimental approach to genetic retinal diseases that may be relevant for beta cells, cardiomyocytes, and neurons [3,4,5]. Optimal donor cells for retinal transplantation are post-mitotic photoreceptor precursors but not mature photoreceptors [28]. Immature photoreceptors generated from ES cells or iPS cells in vitro differentiate through transplantation into the mouse retina [29,30]. In this aspect, the induced retinal cells shown here may be suitable for future cellbased therapy since they are not fully differentiated. Together, the finding contributes substantially to an advance toward cell-based therapy for retinal genetic diseases.

Preparation of tissue and cell culture
Cells were obtained from donors at ages of 10 months, 1 year 8 months and 3 years. Iris tissues were excised from surgical specimens as a therapy for retinoblastoma with the approval of the Ethics Committee of the National Institute for Child and Health Development (NCCHD), Tokyo. The ethics committee of the NCCHD specifically approved this study (approval number, #156). Signed informed consent was obtained from the parents of the donors, and the surgical specimens were irreversibly deidentified. All experiments handling human cells and tissues were performed in line with the Tenets of the Declaration of Helsinki.
The iris was freed from the ciliary body. The iris kept away from a tumor and invasion of retinoblastoma cells were not detected by a pathologist's examination. The iris pieces were cut into smaller pieces and were subjected to explant-culture in the growth medium [Dulbecco's modified Eagle's medium (DMEM)/Nutrient mixture F12 (1:1) supplemented with 10% fetal bovine serum, insulin-transferrin-selenium, and MEM-NEAA (GIBCO)]. Cells derived from the iris pieces were designated as ''iris cells''. ''IPE cells'' were isolated from iris tissues using dispase and trypsin. ''Iris cells'' without ''IPE cells'' were designated as ''iris-stromal (IS) cells'' (Fig. 1A). Ciliary epithelial cells were isolated from pars plana and pars plicata in the same manner as IPE cells. Retinal pieces were cut into smaller pieces and were subjected to explantculture in the same growth medium as iris cells. Second-passage cells were used for all the experiments.

Sequencing of the RB gene
Total RNA was isolated from iris-derived cells of the three donors used in this study. An aliquot of total RNA was reverse transcribed into cDNA. The full-length of RB gene was amplified with Go-Taq polymerase (Promega) using the cDNA. Direct sequencing was performed with a BigDyeH Terminator Cycle Sequencing Kit (Applied Biosystems, Foster City, CA). Sequencing reaction products were run on an automated capillary sequencer (Applied Biosystems 3130xl Genetic Analyzer; Applied Biosystems).

Hanging drop method
Droplets, each of which included 1000 cells in 20 ml of culture medium, were formed on the inverted underside of a single Petri dish cover. The inverted bottom was then set on the top and the entire assembled Petri dish was re-inverted to its normal orientation. The drops of cell suspension were then hanging in the interior of the dish from the inner surface of the cover. The dishes were carefully placed into a 37uC incubator in an atmosphere of 5% CO 2 . Aggregates of cells were allowed to form in the drops for 24 h. Formed aggregates were then transferred to a Poly-D-lysine/laminin-coated 6-well tissue-culture plate (Becton Dickinson) (4 aggregates per well).

Preparation and infection of recombinant retrovirus
The resulting pENTR11-transcription factors were recombined with pMXs-DEST by use of LR recombination reaction as instructed by the manufacturer (Invitrogen). pMXs was a gift from Dr. Kitamura (Tokyo University) and was modified into pMXs-DEST in our laboratory [37]. The retroviral DNAs were then transfected into 293FT cells and three days later the media were collected and concentrated. The iris-derived cells were prepared on laminin-coated six-well dishes or four-well chamber slides and maintained for one day. The cells were infected with abovementioned media containing retroviral vector particles with 8 mg/ ml of polybrene for 5 h at 37uC. After retroviral infection, the media were replaced with the DMEM/F12/B27 medium supplemented with 20 ng/ml bFGF, 40 ng/ml EGF, fibronectin, and 1% FBS. The retrovirus-infected cells were cultured for up to 21 days. We transfected retroviral eGFP under the same conditions to measure efficiency of infection. The frequency of eGFP-positive cells was 90-94% of all cells at 48 h after infection.

Reverse transcriptase-PCR
Total RNA was isolated with an RNeasy Plus mini-kitH (Qiagen, Maryland, USA) or a PicoPure TM RNA Isolation Kit (Arcturus Bioscience, CA, USA) according to the manufacturer's instructions. An aliquot of total RNA was reverse transcribed using an oligo(dT) primer. The design of PCR primer sets is shown in Table 1.

Quantitative RT-PCR
The cDNA templates were amplified (ABI7900HT Sequence Detection System) using the Platinum Quantitative PCR Super-Mix-UDG with ROX (11743-100, Invitrogen). Fluorescence was monitored during every PCR cycle at the annealing step. The authenticity and size of the PCR products were confirmed using a melting curve analysis (using software provided by Applied Biosystems) and a gel analysis. mRNA levels were normalized using G3PDH as a housekeeping gene. The design of PCR primer sets is shown in Table 2.

Light stimulation
A high pressure UV lamp (USH-102D, Ushio) was used as a light source. Diffuse, unpolarized blue and green lights were generated through bandpass filters attached with the fluorescent emission system (BX-FLA, Olympus, Tokyo, Japan). Wavelength of light for stimulation was 460-490 nm for blue and 520-550 nm for green. Duration and timing of light stimulation was monitored by a photodiode (TPS708, Toshiba). Light intensity was calibrated by a light meter (LI-COR, LI-250) that was placed at the focal plane on the stage. To maximize a chance for photoisomelization of photopigment, we applied a strong light to the cell. Light intensity used for stimulation was 390 W/m 2 for blue and 4810 W/m 2 for green.

Electrophysiology
To activate the phototransduction cascade, 11-cis retinal (a gift from the vision research community, the National Eye Institute, National Institutes of Health) was added to the culture medium of the human iris-derived cells to a concentration of 50 mM with  To assess whether a recorded cell had any response to light or not, we used the following criteria: Where I base was an average of holding current for 1 s just before light stimulation and I stim was an average of holding current for 1 s just before the cessation of light stimulation. When I photo was larger than the two times of standard deviation of I base , I photo was judged as a real response to light stimulation. Figure S1 Southern blot analysis. A. Genomic DNA was isolated using the DNeasy kit (Qiagen). Genomic DNA (500 ng) was digested with BamHI restriction enzyme, separated via 0.8% agarose gel electrophoresis, and transferred to Hybond-N membranes (GE Healthcare). The membrane was then fixed under UV irradiation. The full-length RB gene probe was labeled by the AlkPhos Direct Labelling Reagent (GE Healthcare) and hybridized to the blot and detected using CDP-Star detection reagent (GE Healthcare).  Figure S3 Immunocytochemistry using antibodies to rhodopsin (left panels), blue opsin (middle panels) and green/red opsin (right panels) on the cultured iris cells without gene infection (upper panels) and frozen sections of human retina (lower panels). Human iris cells without gene infection (upper panels) and the macular area of human retina (lower panels) served negative and positive controls, respectively, for Fig. 2C. The primary antibodies used were as follows: rhodopsin (goat polyclonal, I-17, Santa Cruz), blue opsin (goat polyclonal, P-13, Santa Cruz), and green/red opsin (goat polyclonal, C-19, Santa Cruz). The secondary antibody used was rabbit polyclonal to goat IgG conjugated with FITC. Nuclei were stained with DAPI. The photoreceptor layer in the retina is positive for rhodopsin, blue opsin and green/red opsin (lower panels from left to right). (DOC) Figure S4 Immunocytochemistry using antibodies to blue opsin (red) and rhodopsin (green). Double-stained immunocytochemistry was performed, by using 2 primary antibodies (mouse monoclonal Ab to rhodopsin and rabbit polyclonal Ab to blue opsin) and 2 secondary antibodies (goat Table 2. Primer sequences for qRT-PCR. anti-mouse polyclonal FITC-labeled Ab and goat anti-rabbit polyclonal rhodamine-labeled Ab). Nuclei were stained with DAPI (blue). Experiments were performed at two weeks after infection. Scale bar represents 50 mm in the rightmost panel.

Supporting Information
(DOC) Figure S5 Electron-microscopic observation. Cells were initially fixed in PBS containing 2.5% glutaraldehyde for 24 h, and were embedded in epoxy resin. Ultrathin sections were doublestained with uranyl acetate and lead citrate, and were viewed under a JEM-1200EX transmission electron microscope (JEOL, Ltd.). After transduction of the RX, CRX, and NEUROD genes into human cultured iris cells, a cilia-associated structure, i.e. centriole (arrow head) surrounded by mitochondria (arrows), was detected.
(DOC) Figure S6 Quantitative analysis of photoreceptor-specific/associated genes expression (blue opsin, s-antigen and recoverin). Individual RNA expression levels were normalized by respective G3PDH expression levels. Vertical axis is relative expression levels of each gene in the siCRX & siNEUROD-transfected cells versus negative control siRNAtransfected cells (%, the mean of relative expression levels of two experiments). siRNA transfection. The cells at 7 days after transduction of the CRX, RX, and NEUROD genes in 6 well plates were transfected with siRNA using Lipofectamine RNAi-MAX Reagent (Invitrogen) according to the protocols recommended by the manufacturer. The cells were harvested 48 h after transfection and analyzed by quantitative RT-PCR. Stealth RNAiTM siRNA Duplex Oligoribonucleotides (siCRX and siNEUROD1, Invitrogen) were used as siRNAs to the CRX and NEUROD genes, and StealthTM RNAi Negative Control Duplexes (Invitrogen) were used as control siRNA. (DOC)