Combinatorial Hedgehog and Mitogen Signaling Promotes the In Vitro Expansion but Not Retinal Differentiation Potential of Retinal Progenitor Cells

Abstract

Purpose.: The in vitro expansion of multilineage competent primary neural progenitor cells is typically limited. Hedgehog (Hh) signaling is required in vivo for the maintenance of stem cell (SC) and progenitor populations in the central nervous system, including the retina. Here we investigated the impact of Hh signaling on in vitro expansion of perinatal mouse retinal progenitor cells (RPCs).

Methods.: Perinatal mouse retinal cells were treated with combinations of Hh agonist (Hh-Ag), epidermal growth factor (EGF)/fibroblast growth factor 2 (FGF2) and the cultures were assayed for long-term growth, gene expression, and dependence on Gli2. Differentiation was assessed in monolayer cultures, following in vivo transplantation and in cellular reaggregates.

Results.: Using a combination of Hh-Ag, EGF, and FGF2, we were able to establish long-term RPC cultures (termed Hh-RPCs). The ability of this combinatorial signaling approach to block quiescence of these was not associated with altered TP53/MDM2 levels or Hh-EGF cooperativity gene expression. Efficient Hh-RPC expansion and monolayer culture establishment requires Gli2, as Hh-RPCs derived from Gli2 knockout retinal tissue fail to generate cultures that can be passaged long-term in vitro. Hedgehog RPCs retain competence for neurogenic and gliogenic differentiation in vitro; however, they fail to engraft and differentiate into retinal cell types following in vivo transplantation to the eye or in vitro when mixed with acutely dissociated perinatal retinal cells.

Conclusions.: Our data show that combining Hh and mitogen signaling is sufficient to promote the expansion of RPCs in vitro, but it is insufficient to maintain competence of these cells for retinal differentiation.

Introduction

The neural retina is an especially tractable system for investigating the regulation of cell diversification in the central nervous system. The six neuronal and one major glial cell type in the adult retina are generated in an invariant, but overlapping, sequence by multipotential retinal progenitor cells (RPCs). ¹ The temporal competence of RPCs to generate the different retinal lineages is regulated largely cell intrinsically, but the final cell repertoire is also impacted by cell extrinsic cues. ² Because it contains a mixture of progenitors and differentiating postmitotic cells, the heterogeneity of the retina presents a challenge for isolating the effects of environment on the developmental progression of progenitors. Moreover, the typically small sample sizes of primary tissue make larger-scale biochemical analysis on purified cell populations difficult. Furthermore, there is a need for efficient generation of large pools of clinically relevant multipotent RPCs for transplantation-based replacement therapies to treat retinal degenerative diseases. Thus, in vitro methods to generate a large pool of purified multipotential RPCs for studies of cell diversification and therapeutics would represent a significant advance in the field. 

Retinal progenitor cells from a number of cell sources, including the fetal retina, adult ciliary margin, and embryonic stem cells, can be induced and/or expanded in culture. ³ Typically, epidermal growth factor (EGF) and fibroblast growth factor 2 (FGF2) are used as the primary mitogens in these culture media, as they have been shown to be mitogenic for rodent RPCs. ^4,5 However, RPC expansion is highly variable. ^6–10 We reasoned that activation of physiologically relevant anti-differentiation pathways would be an effective strategy to improve RPC proliferation in vitro. Sonic Hedgehog (SHH) signaling from ganglion cells (GCs), the projection neurons of the retina, is essential for maintenance of the RPC pool in vivo. ¹¹ Here we show that the combination of Hh agonist (Hh-Ag), EGF, and FGF2 promotes the long-term expansion of undifferentiated mouse RPCs from the perinatal retina. Moreover, efficient Hh-RPC cultures depend on an Hh-Gli2 signaling axis, the major Gli transcription factor required for Hh target gene induction and proliferation in the retina. ¹² Expanded Hh-RPCs retain competence to differentiate into pan-neurons and glia; however, they fail to engraft and differentiate into retinal neurons when transplanted in vivo and when co-cultured with perinatal retinal tissue in vitro. Thus, although supplementation with Hh-Ag and mitogens represents a facile method of long-term expansion of RPCs in vitro, it is not sufficient to maintain competence for retinal lineage differentiation of these cells. 

Materials and Methods

Mice and Retinal Cell Culture

Neonatal C57BL/6 mice were used for in vivo transplantation and as the source of retinal tissue for cell culture. Gli2 ^+/− transgenic mice (obtained from A. Joyner, Sloan-Kettering Institute, New York, NY¹³) were maintained on a CD1 background and littermates were used as controls in analyses involving mutant mice. Transgenic mice were coupled in the late afternoon and the presence of a vaginal plug the next morning was considered as embryonic day 0.5 (E0.5). Animals were cared for and handled according to the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research. For retinal explants, eyes from postnatal day 0 (PN0) or E18.5 mice were dissected in CO₂-independent Dulbecco's modified Eagle's medium (DMEM) (Life Technologies, Carlsbad, CA). The RPE, sclera, and lens were removed. The neural retinas were transferred to a 13-mm polycarbonate filter (0.8 μm pore size; Whatman, Kent, UK), flattened, and cultured with the ganglion cell layer facing up in 24-well plates at 8% CO₂ and 100% humidity in wells containing 0.5 mL SATOS-supplemented serum-free retinal cell culture medium (SFRCM) (DMEM/F12 [1:1], 10 μg/mL insulin, 100 μg/mL transferrin, 100 μg/mL bovine serum albumin [BSA Fraction V], 60 ng/mL progesterone, 16 μg/mL putrescine, 40 ng/mL sodium selenite, 25 μg/mL gentamycin) supplemented with either 20 nM Smoothened agonist (Hh agonist, Hh-Ag, ¹⁴ a kind gift from Curis, Lexington, MA), 25 ng/mL human EGF or 10 ng/mL human FGF2 (Sigma-Aldrich, St. Louis, MO). The concentration of Hh-Ag used in the explant cultures was determined in previous dose response experiments of Hh-Ag–induced proliferation in E18 retinal explants. After 2 days, the retinal explants were digested in 1 mL trypsin solution (0.75 μg/mL; Sigma-Aldrich) at 37°C for 10 minutes. The digestion was stopped by the addition of 1 mL trypsin inhibitor (1 mg/mL in SFRCM) and triturated to single cells. The cell suspension was centrifuged at 200g for 5 minutes and the pellet was resuspended in serum-free stem cell media (SFSCM) (DMEM/F12 [1:1], 6 ng/mL progesterone, 5 ng/mL selenium, 100 μg/mL transferrin, 9.5 μg/mL putrescine, 250 μg/mL insulin, 25 ng/mL human EGF, 10 ng/mL FGF2, 2 μg/mL heparin [Sigma-Aldrich]¹⁵), supplemented with 5 nM Hh-Ag. KAAD Cyclopamine (Millipore, Billerica, MA) was dissolved in dimethyl sulfoxide or ethanol. The cells were cultured in 6-well plates at the density of 5 to 10 × 10⁵ cells per well in 2 mL SFSCM or in 24-well plates at a density of 5 to 10 × 10⁴ cells per well in 0.5 mL SFSCM. The medium was refreshed every 2 or 3 days. After 2 weeks, a monolayer formed and the cultures were passaged every 2 to 3 days by mechanical trituration to obtain single cells that were then diluted 1:3 with fresh culture medium. Differentiated neurons did not survive under these culture conditions and were lost on serial passaging of the cultures. To label Hh-RPCs in S-phase for transplantation assays, BrdU (200 μM) was added to the culture medium for 48 hours before harvesting for transplantation assays. Typically, more than 80% of cells were labeled. Hedgehog RPCs were cloned by plating 100 cells per well in laminin and poly-D-lysine (PDL) (both purchased from Sigma-Aldrich) coated wells of 24-well plates. After 15 days, isolated single spheres were selected and transferred to uncoated wells in 24-well plates, which were then passaged as monolayers. Hedgehog RPCs were nucleofected as per manufacturer's instructions using the basic Nucleofector kit for primary mammalian epithelial cells (Lonza, Basel, Switzerland). Briefly, monolayer Hh-RPCs were mechanically triturated into single cells and 1.5 × 10⁶ cells were mixed with 100 uL nucleofector solution and 5 μg of expression vector encoding GFP under the control of the ubiquitin C promoter. The Hh-RPC/DNA suspension was placed into the nucleofector, program A-033 was run and 500 uL of pre-equilibrated SFSCM was used to gently transfer Hh-RPCs to 6-well plates. 

Pellet Culture

Eyes from C57BL/6 wild-type PN0 mice were dissected in DMEM (Life Technologies) and RPE, sclera, and lens were removed. The neural retinas were digested in 1mL trypsin solution (0.75 μg/mL; Sigma-Aldrich). The digestion was stopped by the addition of 1 mL trypsin inhibitor (1 mg/mL in SFRCM) and triturated into single cells. To produce a pellet, 2 × 10⁶ PN0 cells were mixed in suspension with 1.5 × 10⁴ previously nucleofected Hh-RPCs^GFP+ and centrifuged at 200g for 5 minutes. The resulting aggregated pellet was carefully lifted from the bottom of the centrifuge tube using a 28-gauge insulin syringe (Becton Dickinson, Franklin Lakes, NJ), transferred to a 13-mm polycarbonate filter (0.8 μm pore size; Whatman), and cultured in 24-well plates at 8% CO₂ and 100% humidity in wells containing 0.5 mL SATOS-supplemented SFRCM. After 7 days in vitro (DIV), pellets were processed for immunostaining as described in the Supplementary Methods. 

In Situ Hybridization and IHC

In situ hybridization (ISH) was performed as previously described ¹¹ by using digoxigenin-labeled antisense RNA probes corresponding to mouse Jagged 1, CyclinD1, and Math5. Hedgehog RPCs were grown on glass coverslips and the ISH reactions were performed in 24-well plates. A detailed description of the immunostaining conditions for cells and tissue sections is provided in the Supplementary Methods. The following primary antibodies were used in this study: mouse anti-Nestin (1:200; RDI, Flanders, NJ), mouse anti-CyclinD1 (1:300; Santa Cruz, Dallas, TX), mouse anti-KI67 (1:100; Becton Dickinson), rabbit anti-PCNA (1:500; Santa Cruz), rabbit anti-CyclinD3 (1:100; Santa Cruz), sheep anti-VSX2/CHX10 (1:2000; a kind gift from Rod Bremner, University Health Network, Toronto, ON, Canada), mouse anti-PAX6 (1:6 of tissue culture supernatant; Developmental Studies Hybridoma Bank, Iowa City, IA), rabbit anti-GFAP (1:2000; Sigma-Aldrich), mouse anti-MAP2 (1:200; Synaptic Systems, Goettingen, Germany), rabbit anti-SOX2 (1:1000; Millipore), mouse anti-β3TUBULIN/Tuj1 (1:200; Stem Cell Technology, Vancouver, Canada), mouse anti-GS (1:400; Becton Dickinson), mouse anti-BrdU (1:100; Becton Dickinson), sheep anti-BrdU (1:400; Gene Tex, Hsinchu City, Taiwan), and rabbit anti-Recoverin (1:1000; Millipore). The following secondary antibodies were used: Alexa488 or Alexa568-conjugated goat anti-rabbit or goat anti-mouse antibodies and Alexa 647-conjugated donkey anti-rabbit or donkey anti-mouse IgG (all purchased from Jackson ImmunoResearch, West Grove, PA). Nuclei were stained with Hoechst 33342 or 4′,6′-diamidino-2-phenylindole (DAPI). All sections were viewed under a Zeiss Axioplan microscope and digital images were captured using an Axio Vision 2.05 (Zeiss, Oberkochen, Germany) camera and processed with Adobe Photoshop (Adobe, San Jose, CA). 

Western Blot

Protein was extracted from treated retinal explants or Hh-RPCs using ×1 RIPA buffer (Millipore) with cOmplete Mini EDTA-Free protease inhibitor (Roche, Basel, Switzerland) and total protein was estimated using a Bradford assay (Bio-Rad, Hercules, CA). An amount of 20 μg total protein was run on a 4% to 12% SDS-reducing gradient gel (Bio-Rad), transferred to a nitrocellulose membrane, and probed with mouse anti-MDM2 (1:500; Millipore), mouse anti-p53 (1:3000; Leica Biosystems, Wetzlar, Germany), or mouse anti-glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (1:20,000; Millipore). Donkey anti-mouse IgG horseradish peroxidase (HRP) (1:5000; Millipore) was used as a secondary antibody and protein signals were detected using Luminata Crescendo Western HRP substrate (Millipore). GAPDH was used as a control to confirm equal loading. 

Quantitative Real Time PCR

RNA from Hh-RPCs and cultured PN0 retinas was extracted using Tri-Reagent (Life Technologies) according to manufacturer's instructions. RNA purity and concentration was determined using Ependorph BioPhotometer spectroscopy (Ependorph, Hamburg, Germany). cDNA was then prepared using Moloney murine leukemia virus (Life Technologies) with 2 μg total RNA according to manufacturer's instructions. For quantitative real-time PCR (qRT-PCR), 2 μl of cDNA, 0.8 μl of 2.5 μm forward/2.5 μm reverse primers, and 7.2 μl of HPLC grade water were added to 10 μl of SYBR-Green qPCR master-mix reagent (Sigma-Aldrich). Reactions were performed using an Mx3000P thermocycler (Agilent Technologies, Santa Clara, CA). See Supplementary Table S3 for qRT-PCR primer sequences used in this study. Relative changes in mRNA expression of target genes were determined using the Δ-ΔCt method normalized to Gapdh expression. Statistical significance was determined using a two-tailed Student's t-test. 

In Vivo Transplantation

BrdU-labeled or Nucleofected Hh-RPCs were washed with Hank's Balanced Salt Solution (HBSS; Sigma-Aldrich), centrifuged at 200g for 5 minutes and resuspended in HBSS at a density of 10⁸ cells/mL. Cells were stored on ice and transplanted within 2 hours of harvesting. C57BL/6 recipient mice, aged P2 to P5, were anesthetized with isofluorane/O₂ and injected intravitreally with Hh-RPCs according to the method described by Mizumoto et al. ¹⁶ Briefly, the conjunctiva was cut with small curved scissors to expose the eyeball, a hole was made with a 25-gauge needle at the cornea-sclera junction, and 1 μL of the cell solution (containing 10⁵cells) was injected into vitreous using a 10-μL Hamilton micro syringe (Hamilton Company, Reno, NV) with a 32-gauge blunt needle. After the procedure, the eyelid was closed with a drop of wound adhesive. Only one eye was injected in each mouse and control mice were injected with 1 μL of HBSS. The integration of donor cells was assessed 7 to 15 days after transplantation. 

Results

To investigate whether sustained Hh signaling mediates similar effects on RPCs in vitro as it does in vivo, we compared neuronal differentiation and RPC marker expression in control and Hh-Ag treated retinal explants from postnatal day 1 (PN1) mice. Untreated retinal explants serve as a loss of function model, as they rapidly become Shh deficient because of the death of Shh-expressing GCs, the main source of Hh signaling in the retina. ¹⁷ Exogenous Hh signaling had two major effects on cells in retinal explants. First, compared with control explants, treatment with Hh-Ag for 2 days inhibited neuronal differentiation, as assessed by immunohistochemistry with anti–β3 TUBULIN (Tuj1) antibodies (Fig. 1A). Second, Hh-Ag treatment altered the pattern of PAX6 and CYCLIND3 expression, two RPC markers. The intensity of PAX6 and CYCLIND3 immunoreactivity was increased in Hh-Ag–treated explants compared with the controls (Fig. 1A). Moreover, PAX6⁺ and CYCLIND3⁺ cells were distributed throughout the Hh-Ag–treated explant, compared with the equivalent stage in vivo (PN1), where cells expressing high levels of CYCLIND3 are located in the ciliary body at the periphery of the eyecup and the intensity of PAX6 immunoreactivity in the neuroblast layer is low (Fig. 1A). These short-term effects of Hh-Ag on PAX6 and CYLIND3 expression are not a general consequence of mitogenic stimulation, as they were not observed in explants treated with EGF or FGF2 (Fig. 1A). 

Figure 1

 

Inhibition of neuronal differentiation and rapid induction of progenitor marker expression in Hh-Ag–treated retinal explant cultures. (A) Immunohistochemistry of serial explant sections with antibodies specific for β-TUBULIN (Tuj1, top), PAX6 (second row), CyclinD3 (third row), and DAPI (bottom row) in the PN1 retina and PN1 retinal explants cultured for 2 DIV in serum-free medium under control conditions or in the presence of Hh-Ag, EGF, or FGF2. Note the reduced β-tubulin and the increased Pax6 and CyclinD3 immunoreactivity in Hh-Ag–treated explants compared with control, EGF or FGF2-treated explants. Scale bar: 100 μm. NBL, neuroblast layer; DCL, differentiated cell layer; RGL, retinal ganglion cell layer; CB, ciliary body. (B) Quantification of marker⁺ cells in single-cell dissociates from the PN1 retina (n = 3) and PN1 retinal explants cultured for 2 DIV with Hh-Ag (n = 3). *P < 0.001.

Quantitative analysis of single-cell dissociates from Hh-Ag–treated retinal explants and acutely dissected retina, which represents cells exposed to endogenous SHH signaling, revealed that Hh-Ag treatment in vitro increased the proportions of PAX6⁺ and CYCLIND3⁺ cells by 8- and 35-fold, respectively, compared with the PN1 retina (Fig. 1B). Although Pax6 also is expressed in a subset of retina neurons, we determined that most of the PAX6⁺ cells in Hh-Ag–treated explants were progenitor cells, as they costained with PCNA, a marker of cycling cells (%PCNA⁺PAX6⁺/PAX6⁺ 88.0 ± 2.7 n = 3). There was no significant difference in the proportion of cells expressing cell cycle markers and VSX2 (CHX10) in the Hh-Ag–treated explants and the PN1 retina (Fig. 1B). Thus, the enrichment for CYCLIND3 and PAX6 expression in explants following in vitro exposure to Hh-Ag likely reflects the induction of these markers in the explants as opposed to the expansion or altered survival of a distinct subset of cells in the explants. 

The enrichment of PAX6⁺ and CYCLIND3⁺ cells in Hh-Ag–treated explants prompted us to investigate the effect of sustained Hh pathway activation on the growth of RPCs outside of the explant environment. Explants were primed for 2 days with Hh-Ag, dissociated into single cells, and cultured in serum-free medium supplemented with Hh-Ag, EGF, and FGF2 or combinations of single and double growth factors (Fig. 2A). Only cultures supplemented with the combination of Hh-Ag, EGF, and FGF2 generated monolayer cultures that could be passaged routinely after 6 weeks (Figs. 2A, 2B, and Table) and we refer to these monolayer cultures as Hh-RPCs. Consistent with previous reports, ^18,19 we could establish monolayer cultures of RPCs by growing acutely dissociated retinal cells in medium supplemented with EGF/FGF2 (referred to as “mitogens”); however, this approach was not as reliable as the Hh-RPC method (Table); the cultures had a high tendency to quiesce (Fig. 2C) or to develop aggregates (Fig. 2D) by 14 days and could rarely be passaged after 6 weeks (Table). Treatment with the Hh pathway antagonist, cyclopamine, prevented monolayer formation and inhibited proliferation in retinal explants, confirming the specificity of the Hh-Ag in this context (Supplementary Fig. S1). 

Figure 2

 

Propagation of RPCs in the presence of Hh-Ag and mitogens. (A) Schematic of method to generate Hh-RPCs. P0 eyes are removed, the lens and retinal pigment epithelium dissected away from the retina, and the tissue is placed on a polycarbonate filter in the presence of Hh-Ag for 2 DIV followed by dissociation into single cells and further culturing for 14 DIV in the presence of EGF, FGF2, and Hh-Ag. (B, C) Phase-contrast images of dissociated cell cultures from mouse retinal explants treated for 2 days with Hh-Ag followed by dissociation and culture for 11 days with Hh-Ag, EGF, and FGF2 (B) or EGF alone (C). (D) An example of the cellular aggregates that form in RPC cultures generated by direct dissociation of retina followed by monolayer culture in EGF and FGF2. (E) In situ hybridization for Math5, Jagged-1, and CyclinD1 mRNA in Hh-RPCs. (F) Comparison of marker expression in Hh-RPCs after culture for 18 days and following cryopreservation and culture for an additional 8 days. Scale bar: 50 μm.

Table

 

Comparison of the Effects of Growth Factor and Culture Conditions in the Establishment of RPC Monolayer Cultures

Culture Condition	No. of Trials	No. of Trials That Resulted in the Successful Establishment of RPC Monolayers*
Direct dissociation†
Supplementation with EGF+FGF2	9	2
Supplementation with Hh-Ag+EGF+FGF2	4	3‡
2DIV explant step followed by dissociated cell culture
Explant: Hh-Ag	20	20
Dissociated culture: Hh-Ag+EGF+FGF2
Explant: Hh-Ag	10	0
Dissociated culture: EGF + FGF2
Explant: Hh-Ag	2	0
Dissociated culture: Hh-Ag+EGF
Explant: Hh-Ag	2	0
Dissociated culture: Hh-Ag+FGF2
Explant: Hh-Ag	2	0
Dissociated culture: Hh-Ag
Explant: Hh-Ag	2	0
Dissociated culture: EGF
Explant: Hh-Ag	2	0
Dissociated culture: FGF2
Explant: Hh-Ag	2	0
Dissociated culture: no growth factors
Explant: no Hh-Ag	2	0
Dissociated culture: Hh-Ag + EGF + FGF2
Explant: no Hh-Ag	2	0
Dissociated culture: EGF + FGF2

*  Success was determined by the establishment of a monolayer of cells with neuroepithelial morphology after 2 weeks in culture. A trial was considered a failure if there was no cell growth after 2 weeks or if cellular aggregates (spheres) formed in the cultures; in our experience, the formation of aggregates was always associated with growth cessation a few weeks later.

†  Retinas were dissected, enzymatically dissociated to single cells, and plated on tissue culture plastic in the presence of the indicated growth factors.

‡  One of these cultures could not be propagated after the 3-month mark, the other two were not passaged past 14 days.

Hedgehog RPCs exhibited a neuroepithelial morphology and, based on immunohistochemistry, RT-PCR and microarray analysis expressed markers characteristic undifferentiated neural progenitor and stem cells, including Jagged1, CyclinD1 (Fig. 2E), CYCLIND3, Nestin, PAX6, VSX2, SOX2, Glutamine synthetase (GS), Hes1, Notch1, retinal progenitor markers Vsx2/Chx10, Lhx2, Six3, Six6 (Fig. 2F, Supplementary Figs. S2A, S2B, Supplementary Table S1), proneural genes, including NeuroD1 and Mash1 (Supplementary Fig. S2B), but not neurogenic genes or terminal differentiation markers (Fig. 2F, Supplementary Fig. S2B, and data not shown). Because they were derived from perinatal retinal tissue, Hh-RPCs also could be Müller glia, which is consistent with expression of ApoE²⁰ and S100β²¹ (Supplementary Table S1), and a pattern of Hh target gene expression (Ptch1⁺Ptch2⁺Gli1⁻) that is typical of adult Müller glia²² (Supplementary Fig. 2B). However, they did not express the full complement of adult Müller glial markers, notably CRALBP/RLBP1, suggesting that if they are Müller they are incompletely differentiated (data not shown). We could maintain Hh-RPCs in culture for more than 1.5 years (the longest continuous culture that we attempted), and recover them following cryopreservation without significant changes in marker expression (Fig. 2F). Hedgehog RPCs were not transformed cells, as they required continuous growth factor supplementation for survival and they did not form tumors after in vivo injection to the eye (n = 41 in vivo transplantations to PN2-PN4 host analyzed 1 to 4 weeks post transplantation). Thus, Hh-Ag priming in explants followed by dissociated cell culture in the presence of Hh-Ag, EGF, and FGF2 is a reliable method for establishing cultures of neuroepithelial-like cells that are stable for long-term passage in vitro. 

Next, we sought to determine the molecular characteristics that distinguish Hh-RPCs from monolayers that were exposed to mitogens alone or mitogens and Hh-Ag. In the cerebellum, Hh mediates granule neuron progenitor proliferation; in part, through Mdm2-dependent inhibition of Tp53. ²³ Thus, we investigated whether a similar Hh-dependent Mdm2-Tp53 axis was responsible for the long-term culture of Hh-RPCs in vitro. Retinal progenitor cell cultures were established, as described above, and supplemented with Hh-Ag and mitogens or mitogens only at the monolayer stage of the protocol. After 7 days in vitro, TP53 and MDM2 levels were compared via western blot. Because cultures supplemented with mitogens alone are frequently quiescent by 14 days, we reasoned that molecular differences between the different growth factor supplementations would be apparent at the midway point of the culture period. Surprisingly, growth factor supplementation did not affect MDM2 levels, and TP53 was undetectable in both conditions (Fig. 3A and data not shown). We next addressed whether the efficient propagation of Hh-RPCs was associated with differences in TP53/MDM2 levels at the explant priming stage. PN0 explants were stimulated with Hh-Ag, EGF, or a combination of Hh-Ag and EGF for 2 days in vitro and analyzed for TP53 and MDM2 by western blot. Similar to Hh-RPC cultures, TP53 protein was undetectable in retinal explants (Fig. 3B). Interestingly, MDM2 levels in explants were increased to a similar extent by Hh-Ag or mitogen treatment (Fig. 3B), suggesting that mitogen stimulation, but not specifically Hh activation, promotes MDM2 expression in explants. 

Figure 3

 

The long-term propagation property of Hh-RPCs does not correlate with TP53/MDM2 protein levels or differences in Hh-EGF cooperative response genes at monolayer and explant stages of culture. (A) Western blot with lysate from Hh-RPCs treated with the indicated mitogens for 7 days post dissociation or acutely dissected P0 retina; 25 μg total protein was run on a 4% to 15% SDS gel and probed with the indicated antibodies. (B) Western blot with lysate from acutely dissected PN0 retinal explants were cultured for 2 DIV in the presence of the indicated mitogen and/or Hh-Ag; 25 μg of total protein was run on a 4% to 15% SDS gel and probed with the indicated antibodies, and 5 μg UV-exposed COS cell lysate was used as a positive control for TP53. (C) Hh-EGF signaling cooperative response gene expression in Hh-RPCs 7 days post dissociation measured using qRT-PCR. *P < 0.05. (D) Hh-EGF signaling cooperative response gene expression in 2 DIV–treated retinal explants measured using qRT-PCR. *P < 0.05.

Recently, synergistic Hh-EGF signaling has been shown to determine the oncogenic potential of basal carcinoma and pancreatic cancer initiation, in part through enhanced expression of Sox2, Sox9, Jun, Cxcr4, and Fgf19, termed “cooperation response genes.” ²⁴ We hypothesized that synergistic Hh-mitogenic signaling may promote long-term propagation of Hh-RPC in vitro through enhanced cooperation response gene expression. To test this hypothesis, we compared cooperation gene expression by qRT-PCR in RPC monolayer cultures and in retinal explants. We observed no significant difference in Hh-EGF cooperation response gene expression between RPC cultures treated with Hh-Ag and mitogens or mitogens alone after 7 days, or in explants cultured for 2 days in the presence of Hh-Ag or EGF (Figs. 3C, 3D). Taken together, the long-term propagation property of Hh-RPCs is not correlated with altered expression of Hh-EGF cooperation response genes in explants or differences in TP53/MDM2 protein levels at the monolayer or explant stages. 

The transcriptional output of the Hh pathway is mediated by the combinatorial activity of Gli1, Gli2, and Gli3 Zinc finger transcription factors. ²⁵ Maximal Hh-mediated progenitor proliferation and target gene induction in the retina is dependent on Gli2 ¹² and, therefore, we reasoned that Gli2 may also have a similar role in the in vitro expansion of Hh-RPCs. We generated RPC cultures from Gli2 knockout (Gli2KO), heterozygote (Gli2HET) or wild-type (Gli2WT) littermates and observed that Gli2KO cultures consistently failed to establish a monolayer by 14 days in vitro (Figs. 4A, 4A''). Cell quantification at 14 days confirmed that there was a significant decrease in the total cell number in cultures established from Gli2KO retinae, compared with Gli2WT or Gli2HET (Fig. 4B), suggesting that Gli2 is required for the efficient establishment of Hh-RPCs in vitro. 

Figure 4

 

Gli2 is required for the propagation and outgrowth of Hh-RPCs. Hedgehog RPC cultures were established from the retinas of E18.5 mice of the indicated genotypes. Acutely dissected E18.5 Gli2 null retinae were individually dissected and cultured for 2 DIV in the presence of Hh-Ag. After 2 DIV, retinal explants were individually dissociated, counted, and plated at 5 × 10⁴ cells per well and treated with Hh-Ag, EGF, and FGF2. After 14 days, total live cell numbers were counted in duplicate. Representative pictures of (A) Gli2WT, (A') Gli2HET, and (A'') Gli2KO Hh-RPC cultures at 14 days post dissociation. (B) Quantification of total live cell numbers. Sample size indicated on bars in graph. *P < 0.05.

To investigate the competence of Hh-RPCs to differentiate into retinal neurons and glia, we exposed Hh-RPCs to standard monolayer-based differentiation conditions that lack mitogens and contain low levels of serum ¹⁵ and stained the cells with cell-type–specific markers. Hedgehog RPCs differentiated into cells that expressed pan neuron and astrocytic cell markers and this neurogenic potential required continuous exposure to Hh-Ag under growth conditions (Fig. 5). However, these conditions did not promote the differentiation of Hh-RPCs to retinal neurons or adult Müller glia, as we were unable to detect the expression of retinal cell-type–specific markers, including Rhodopsin (rods), Recoverin (photoreceptors, subset of bipolars), Syntaxin (amacrine), PKC (rod bipolars) and Calbindin (subset of amacrines, horizontal cells), and CRALBP/RLBP1 (Müller glia) (data not shown). Retinal progenitor cell differentiation is context dependent ²⁶ ; therefore, we monitored the differentiation of Hh-RPCs following exposure to a retinal environment. BrdU-labeled Hh-RPCs were transplanted into the vitreous of PN3 mouse eyes and after 14 days in vivo the tissues were harvested and processed for immunohistochemistry (IHC) for BrdU and cell-type–specific markers. In contrast with previous in vivo transplantation studies of mitogen-expanded rodent RPCs or Müller glia to the perinatal retina, ^27,28 we observed extensive BrdU staining in the inner and outer nuclear layers of the retina that colabeled with every retinal cell-type–specific marker, with the exception of ganglion cell (GC) markers, in the correct location and morphology for that lineage (data not shown). 

Figure 5

 

Continuous Hh pathway stimulation promotes the neurogenic competence of Hh-RPCs. (A) Analysis of Hh-RPCs grown in differentiation conditions and immunostained for MAP2, a neuronal marker, and GFAP, an astrocyte marker. Hedgehog RPCs that were cultured for 31 days were cultured for an additional 2 weeks with or without Hh-Ag in medium supplemented with EGF and FGF2. To assay for differentiation, the cells were transferred to wells coated with PDL+ laminin and cultured in medium containing 1% fetal calf serum in the absence of mitogens or Hh-Ag for 7 days. Scale bar: 50 μm. (B) Quantification of neurons and glia follow differentiation of Hh-RPPs that were grown for 2 weeks before differentiation in medium with and without Hh-Ag.

Because a similar degree of cell integration of transplanted cultured RPCs marked with other lineage tracers has not been reported previously, ^27–30 we were prompted to consider the possibility that the BrdU labeling in Hh-RPC–transplanted retina represents BrdU incorporation of host cells from unincorporated BrdU in the injection solution or transfer from donor cells. BrdU transfer from dead transplanted cells has been reported, ^31,32 and the PN3 retina contains dividing progenitors, which could incorporate the label. To determine if BrdU from dead Hh-RPCs can be incorporated into host retinal tissue in vivo, BrdU-labeled Hh-RPCs were repeatedly freeze-thawed before injection, and loss of viability was confirmed by trypan blue staining (trypan blue+ = dead Hh-RPCs). Live and dead Hh-RPCs and a cell-free preparation, used as a control for residual BrdU carryover from the cell media, were injected into PN3 mouse eyes and processed for IHC 7 days post surgery. BrdU labeling was not observed in retinae injected with the cell-free preparation, ruling out any residual carryover of BrdU from labeling in vitro. Surprisingly, retinas transplanted with dead or live Hh-RPCs exhibited a similar high degree of BrdU labeling (Fig. 6), demonstrating that host tissue can incorporate BrdU from dying RPCs and that BrdU is not an adequate label to track transplanted Hh-RPCs. 

Figure 6

 

Dead Hh-RPCs transplanted in vivo transfer the BrdU label to the host retinal tissue. 10⁵ BrdU-labeled dead Hh-RPCs (A, A') or live RPCs (B, B') were injected into the vitreous of P3 mice. After 7 days, the retinae were isolated, processed, and stained with antibodies specific for BrdU (A, B) and a nuclear marker (Hoechst) (A', B'). BrdU-labeled E14.5 wild-type retina (C) was used as a positive control. BrdU staining in eyes injected with a cell-free preparation (D) as a control for carryover of unincorporated BrdU. Boxed areas represent ×1.5 zoom of (A–D). Scale bar: 100 μm.

As an alternative cell tracing method to track Hh-RPCs, we attempted to transfect Hh-RPCs with replication incompetent retrovirus expressing GFP, but Hh-RPCs could not be infected efficiently (data not shown). Therefore, we used nucleofection ³³ to transiently transfect Hh-RPCs with an expression vector encoding GFP under the control of the ubiquitin C promoter (33% GFP+/DAPI+ cells 24 hours post nucleofection, Hh-RPC^GFP+). Following subretinal transplantation into PN3 mouse eyes, injected Hh-RPC^GFP+ cells were observed at the site of injection but not within the host retina (data not shown). To circumvent the limited engraftment of these cells, we adopted a pellet culture method in which Hh-RPCs^GFP+ were mixed with dissociated PN0 retinal cells, centrifuged into pellets, and cultured for 7 days. Pellet cultures are an established method for monitoring retinal differentiation in vitro, ³⁴ and the direct mixing of Hh-RPCs and unlabeled retinal cells bypasses the integration step associated with transplantation, thereby allowing us to assay the differentiation potential of Hh-RPCs exposed to a retinal environment. After 7 days post mixing, the pellets were harvested and processed for IHC for GFP and cell-type–specific markers. Hedgehog RPCs^GFP+ were present throughout the pellet tissue and exhibited complex neuron-like morphology, but did not express markers of photoreceptors or glia (Fig. 7A and data not shown). Although pellet cultures of retinal cells are not laminated, the cells are arranged into distinct rosettes that consist of inner (amacrine, bipolar, Müller) and outer (photoreceptor) nuclear layer retinal cells. ³⁴ Hh-RPCs^GFP+ did not appear to be incorporated into rosettes (Fig. 7A) or to express photoreceptor (Recoverin) (Fig. 7B) or glial (GFAP) (Fig. 7C) markers, and instead were more commonly located in more acellular regions of the pellet (lumens of rosettes) or in close proximity to the polycarbonate culture filter (Fig. 7). The absence of GFAP expression in Hh-RPC^GFP+ cells in pellets could reflect an inhibitory effect of this environment on GFAP induction and/or poor survival of GFAP+ derivatives of these cells. Taken together, although in vitro expanded Hh-RPCs do not express markers of differentiated retinal neurons or glia, they retain competence for pan neuron and glia differentiation in vitro that does not depend on signaling from differentiating retinal cells. 

Figure 7

 

Hedgehog RPCs cultured in aggregates with neonatal wild-type retinal cells adopt neuron-like morphology but do not express retinal cell markers. An amount of 2 × 10⁴ GFP-transfected Hh-RPCs were mixed with 2 × 10⁶ acutely dissociated P0 wild-type retinal cells and aggregated by gentle centrifugation. After 7 days, in vitro pellets were fixed, cryosectioned, and stained with a nuclear marker (DAPI) (A'–C') or antibodies specific for GFP, marking transfected Hh-RPCs (A–C, A'–C'), photoreceptors (anti-Recoverin) (B, B'), and activated Müller glia and astrocytes (anti-GFAP) (C, C'). Aggregated Hh-RPCs show variable branching patterns (indicated by arrowheads) (A, A') similar to neurons, but do not participate in rosette structures composed of photoreceptors (indicated by dotted circles) (A, A', B, B'). Scale bar: 100 μm.

Discussion

Here we show that exposing retinal explants to Hh-Ag followed by dissociated cell culture in the presence of Hh-Ag and mitogens is a reliable approach for the establishment of monolayer cultures of RPCs that can be passaged long-term, and this effect appears to be dependent on Gli2 levels. Hedgehog RPCs can give rise to generic neurons and glia in low serum-differentiation conditions, but fail to give rise to retinal-specific cell types, including photoreceptors, even when cultured in close proximity to retinal cells in a pellet culture. Our work also highlights the pitfalls of using BrdU to label transplanted cells into the retina, as we found that host tissue readily uptakes BrdU from injected cells, likely as a result of transfer from dead cells. 

Synergy between Hh and EGF signaling has been reported in neural progenitor development in other brain regions, ³⁵ and is mediated through the p53 pathway, ²³ expression of Hh-EGF cooperation genes, ³⁶ and stabilization of GLI2. ³⁷ Although the expansion of Hh-RPCs was not associated with changes in TP53 or cooperation gene expression, it did require Gli2. In RPCs, Gli2 is required for induction of CyclinD1 and Hes1, two key effector genes of Hh-induced proliferation in the mouse retina, ¹² an effect that is not mimicked by EGF or FGF2 treatment in retinal explants (Wall DS, Wallace VA, 2008, unpublished observations). ^11,12 The facile expansion of RPCs in response to combinatorial Hh-Ag and mitogen stimulation could, therefore, reflect a Gli2-dependent requirement for simultaneous activation of self-renewal and cell-cycle target genes for the maintenance RPC self-renewal. Sustained Notch signaling is also associated with persistence of progenitor cells with stem cell–like characteristics in the murine retina; however, the neurogenic competence of these cells has not been reported. ³⁸  

Although we show that Hh-RPCs with neurogenic and gliogenic potential can be maintained in vitro, they do not differentiate into photoreceptors. This observation is consistent with a recent report describing the lack of retinal differentiation capacity of EGF/FGF-expanded mouse RPCs, ⁹ which was associated with the downregulation of expression of retinal identity genes. In contrast, the expression of several retinal identity genes was maintained in Hh-RPCs, including Vsx2, Pax6, Lhx2, Six3, and Six6. Although Hh-RPCs express several RPC and retinal identity markers, they could represent Müller glia, a cell type that exhibits neurogenic potential ³⁹ and expresses several RPC markers. ⁴⁰ A Müller glial cell origin for Hh-RPCs is consistent with the observation that Hh pathway activation promotes Müller glial cell development in retinal cultures, ^41,42 and upregulates expression of markers of activated Müller glia, including CyclinD3, Pax6, and Vsx2. ⁴³ Moreover, Hh-RPCs express Müller markers, and exhibit an adult Müller cell–specific pattern of Hh target gene expression. ⁴⁴ Hh-RPCs are, however, distinct from mature Müller glia because they do not express the full complement of Müller glia markers and they express Jagged1, which is normally restricted to non-neurogenic cells in the periphery of the retina. Irrespective of their classification as RPC or Müller glia, our findings, together with those of Czekaj et al., ⁹ suggest that culture conditions affect the ability of these cells to differentiate into photoreceptors. One possibility is that the monolayer culture conditions could promote the respecification of Hh-RPCs to cell types, such as hypothalamic progenitors, which are induced by higher levels of Hh signaling at the midline at the same rostral level as the eye field. ⁴⁵ However, we failed to detect the expression of several hypothalamic markers in cultured Hh-RPCs. 

Studies in animal models support the feasibility of cell-based transplantation strategies to restore vision in the diseased retina ^30,46 ; however, the development of efficient and reliable methods for the in vitro expansion of purified RPCs from a variety of stem cell sources is a significant barrier to clinical applications of this technology for the treatment of retinal degenerative diseases, such as AMD. Mimicking physiologically relevant growth and cell fate signaling is a useful strategy for directing the differentiation of region-specific neural progenitors and neuron subtypes from stem cells in vitro. ^47,48 We reasoned that combining physiologically relevant mitogenic and antidifferentiation signals would promote the ex vivo expansion of multipotential RPCs. Although this strategy supports the expansion of a “generic”-type of neural progenitor from the retina, it is not sufficient to maintain retina differentiation potential and suggests that to be successful, additional factors and intercellular interactions need to be considered in these in vitro manipulations. 

Supplementary Materials

pdf S2 

tif S1 

xlsx S3 

Acknowledgments

We thank Rod Bremner and Jack Saari for antibodies; Alex Joyner for mice; Curis, Inc., for Smoothened agonist; and Rashmi Kothary, Dave Picketts, and Rod Bremner for comments on the manuscript. We thank Chantal Mazerolle, Sherry Thurig, and Andrew Ha for assistance with the transplantation experiments. 

Supported by operating grants to VAW from the Canadian Institute of Health Research, Foundation Fighting Blindness Canada, and Stem Cell Network of Canada. 

Disclosure: R. Ringuette, None; Y. Wang, None; M. Atkins, None; A.J. Mears, None; K. Yan, None; V.A. Wallace, None