|Home | About | Journals | Submit | Contact Us | Français|
Cell extrinsic signals can profoundly influence the production of various neurons from common progenitors. Yet mechanisms by which extrinsic signals coordinate progenitor cell proliferation, cell cycle exit, and cell fate choices are not well understood. Here, we address whether Hedgehog (Hh) signals independently regulate progenitor proliferation and neuronal fate decisions in the embryonic mouse retina. Conditional ablation of the essential Hh signaling component Smoothened (Smo) in proliferating progenitors, rather than in nascent postmitotic neurons, leads to a dramatic increase of retinal ganglion cells (RGCs) and a mild increase of cone photoreceptor precursors without significantly affecting other early born neuronal cell types. In addition, Smo deficient progenitors exhibit aberrant expression of cell cycle regulators and delayed G1/S transition, especially during the late embryonic stages, resulting in a reduced progenitor pool by birth. Deficiency in Smo function also causes reduced expression of the basic helix-loop-helix transcription repressor Hes1 and preferential elevation of the proneural gene Math5. In Smo and Math5 double knockout mutants, the enhanced RGC production observed in Smo deficient retinas is abolished, whereas defects in the G1/S transition persist, suggesting that Math5 mediates the Hh effect on neuronal fate specification but not on cell proliferation. These findings demonstrate that Hh signals regulate progenitor pool expansion primarily by promoting cell cycle progression and influence cell cycle exit and neuronal fates by controlling specific proneural genes. Together, these distinct cellular effects of Hh signaling in neural progenitor cells coordinate a balanced production of diverse neuronal cell types.
The mature vertebrate retina consists of seven specific neuronal cell types uniquely devoted to sensing and processing visual information. Retinal ontogeny follows an evolutionarily conserved order, with retinal ganglion cells (RGCs), horizontal cells, cone photoreceptors, and amacrine cells first becoming postmitotic, followed by a late wave of cell birth giving rise to rod photoreceptors, bipolar cells, and Muller glia (Altshuler et al., 1991; Young, 1985a; Young, 1985b; Spence and Robson, 1989). Cell lineage tracing and molecular genetic analyses have revealed that the interplay between cell-extrinsic cues and cell-intrinsic factors is critical for the formation of a functional retinal network (Turner and Cepko, 1987; Holt et al., 1988; Wetts and Fraser, 1988; Lillien, 1995; Vetter and Brown, 2001; Mu et al., 2005; Viczian et al., 2003; Furukawa et al., 1997b; Mears et al., 2001; Pan et al., 2008; Ohsawa and Kageyama, 2008). However, the precise mechanisms by which extracellular signals coordinate progenitor cell behaviors are not well understood.
The hedgehog (Hh) family of molecules affects multiple aspects of retinogenesis. In early neurogensis, Sonic hedgehog (Shh) derived from first-born RGCs promotes propagation of the neurogenic wave front (Neumann and Nuesslein-Volhard, 2000), but suppresses RGC genesis as these neurons accumulate (Zhang and Yang, 2001; Yang, 2004; Wang et al., 2005). Shh signals also appear to influence the growth and trajectory of RGC axons (Kolpak et al., 2005; Sánchez-Camacho and Bovolenta, 2008). In zebrafish, reduction of Hh activities affects differentiation of late cell types including Muller glia, bipolar cells, GABAergic amacrine cells, and photoreceptors (Shkumatava et al., 2004; Stenkamp and Frey, 2003). Furthermore, laminar organization of the retina is disrupted in Shh mutants (Shkumatava et al., 2004; Wang et al., 2002).
Despite the established function of Shh as a mitogen in several compartments of the central nervous system (reviewed by Ruiz I Altaba et al., 2002; Cayuso et al., 2006), the precise role of Hh in retinal proliferation remains controversial. In rodents, recombinant Shh-N promotes retinal progenitor proliferation in cultures (Jensen and Wallace, 1997; Levine et al., 1997), and partial depletion of Shh decreases proliferation (Wang et al., 2005). In Xenopus, inhibiting Hh signals hinders cell cycle progression (Locker et al., 2006; Agathocleous et al., 2007). However, in zebrafish, Hh mutation appears to cause a prolonged period of cell proliferation (Shkumatava and Neumann, 2005).
In this study, we have examined mechanisms through which Hh signals affect important behaviors of neural progenitors, by analyzing conditional mutants of the essential Hh signaling component Smoothened (Smo) (Alcedo et al., 1996; van den Heuvel and Ingham, 1996) and double mutants of Smo and the proneural gene Math5 (Kanekar et al., 1997; Brown et al., 1998; Wang et al., 2001; Brown et al., 2001). We provide conclusive evidence that Hh signals profoundly influence progenitor cell proliferation and affect fate decisions of specific neuronal types. Furthermore, we show that distinct Hh signaling effects are mediated by different intracellular machineries during the neurogenic cell cycle. These findings thus provide mechanistic insights on how cell-extrinsic cues coordinate neural network formation.
Mice carrying the floxed Smo allele (Smoflox, Long et al., 2001) and the Chx10-Cre transgene (Rowan and Cepko, 2004) were obtained from the Jackson Laboratory (Bar Harbor, ME). Mice encoding Cre in the Math5 locus (Math5Cre-ki) were previously described (Yang et al., 2003). All three lines were backcrossed into the C57BL/6J background for a minimum of 3 generations. To generate Smo conditional knockout mice, homozygous Smoflox/flox females were crossed with either Smoflox/+; Chx10-Cre or Smoflox/+; Math5Cre-ki/+ male mice. To generate Smo and Math5 double knockout mice, Smoflox/flox; Math5Cre-ki/Cre-ki females were crossed with Smoflox/+; Math5Cre-ki/+; Chx10-Cre male mice. Genotypes were determined by PCR using primers listed in Suppl. Table 1. Animal procedures were approved by University of California Los Angeles Animal Research Committee.
Immunolabeling was performed as previously described (Zhang and Yang, 2001; Hashimoto et al., 2006). Retinal cryosections fixed in 4 % paraformaldehyde were incubated with the following primary antibodies against: Brn3a (1:100; Chemicon), GFP (1:500; Molecular Probes), β-tubulin (1:800; Covance), GγC (1:1000, Cytosignal), Pax6 (1:1000, Chemicon), AP2α (1:4; DSHB), NF145 (1:1000; Chemicon), and 5-bromo-2-deoxyuridine (BrdU) (1:100; Abcam). Secondary antibodies conjugated with Alexa 488 or Alexa 594 (1:500; Molecular Probes) were used. In vivo BrdU incorporation was conducted by introperitoneal injection of 2.5 mg BrdU per female 24 hours before animals were sacrificed. Sections were treated with 4N HCl for 5 minutes prior to incubation with anti-BrdU antibody. Immunofluorescent images were captured using a Nikon E800 microscope or a laser scanning confocal microscope (Leica TCS-SP).
In situ hybridization was performed using Digoxigenin-labeled RNA probes as previously described (Yang and Cepko, 1996). Mouse Otx2, Crx, Ngn2, Math3 and Cyclin D1 cDNAs were generated by reverse transcription followed by PCR using primers listed in Suppl. Table 2 and authenticated by DNA sequencing. Math5 and Gli1 cDNAs were previously described (Wang et al., 2001; Sasaki et al., 1999).
Dissected retinas were dissociated with 0.1 % trypsin (Sigma) in calcium magnesium free (CMF) HBSS (GIBCO) for 15 minutes at 37°C. Dissociated cells were fixed in 0.25 % paraformaldehyde in HBSS for 30 minutes at room temperature followed by incubation with 0.1 % Triton-X100 in HBSS. Antibodies were diluted in CMF HBSS containing 1 % fetal calf serum (FCS), 0.2 % goat serum, 0.2 % donkey serum and 0.1 % Triton-X100. For BrdU staining, cells were first treated with 0.2 N HCl and washed once with CMF HBSS. Cells were incubated for 60 minutes at room temperature with the following primary antibodies against: PCNA (1:100, Sigma), GFP (1:500; Molecular Probes), NF145 (1:1000; Chemicon), Brn3a (1:100; Chemicon), βTubulin (1:800; Covance), Crx (1:100; Abnova), GγC (1:1000; Cytosignal), AP2α (1:4; DSHB), calbindin (1:500; Chemicon), Cyclin D1 (1:100; Chemicon), Lim1 (1:200, Poch et al, 2007), p27Kip1 (1:100; BD Bioscience), p57Kip2 (1:20; Santa Cruz Biotech), and BrdU (1:2; Amersham). Cells were then washed once with HBSS containing 0.1 % Triton-X100 and incubated for 30 minutes at room temperature with 10 μg/ml 4′, 6-diamidino-2-phenylindole (DAPI; Roche) and Alexa-488 or Alexa-647 conjugated secondary antibodies. Flow cytometry was performed using an LSR flowcytometer (BD Biosciences) and analyzed with Cell Quest (BD), Modfit (BD), and FlowJo (Tree Star) software.
Retinal explants were pulse-labeled with 25 μM BrdU in basal medium (10 mM HEPES pH 7.0 in DMEM and F12 at a 1:1 ratio) containing 5 % FCS and penicillin/streptomycin, and then transferred onto 0.4 μm pore size Minicell membrane (Milipore) and cultured for 0-21 hours at 37°C in 5 % CO2.
Each sample combined multiple E15.5 retinas according to the genotypes, and total RNAs were purified by ISOGEN (Nippongene, Japan). Single stranded cDNAs were prepared using Superscript III (Invitrogen). The cDNA derived from 10 ng RNA and 100 nM of each primer were used for a single reaction. The reactions were carried out for 40 cycles with annealing at 60-62°C for 15 seconds followed by extension at 95°C for 1 minute using real-time PCR master mix containing SYBR Green (Applied Bioscience Inc.) and StepOne real-time PCR system (Applied Bioscience Inc.). Real-time PCR data were normalized with 18S ribosomal RNA. The primers used were either designed by using Primer3 (http://frodo.wi.mit.edu/) or selected from a primer bank (http://pga.mgh.harvard.edu/primerbank/index.html)(Suppl. Table 2).
For quantification of cells plated and immunolabeled as a monolayer, a minimum of 300 and up to 1000 cells per sample was quantified using ImagePro PLUS software (Media Cybernetics). For flow cytometry, a minimum of 30,000 cells was analyzed per sample. Data were presented as mean ± s.e.m. For all pairwise analyses, the Student t test was used. For comparison of multiple sample groups, ANOVA analyses followed by Tukey-Kramer test were performed (Hsu, 1996). P values less than 0.05 were considered statistically significant.
To completely eliminate the cellular response to Hh signals, we performed retina-specific ablation of the Smo gene, which is essential for signal transduction of all known Hh ligands (Alcedo et al., 1996;van den Heuvel and Ingham, 1996). To determine the temporal requirement for Hh signaling, mice encoding Smo.floxed alleles (Smoflox, Long et al., 2001) were first crossed with a mouse line in which the Cre gene had replaced the Math5 gene (Math5cre-ki, Yang et al., 2003). Math5 expression is normally found in a subset of progenitor cells poised to exit the mitotic cell cycle and in nascent postmitotic RGCs as early as embryonic day 11.5 (E11.5)(Brown et al., 1998; Wang et al., 2001; Le et al., 2006). Test crosses between Math5cre-ki and ROSA26R lacZ Cre reporter mice (Soriano, 1999) showd that nearly all of β-Gal positive cells resided in the retinal RGC layer at E15.5 (data not shown). Thus, Cre function in Math5cre-ki mice was active in cells that were either exiting or had exited the cell cycle to give rise to postmitotic RGCs. Further analyses revealed no difference in embryonic neuronal production among the control (Smoflox/flox; Math5+/+), Smo heterozygous (Smoflox/+; Math5cre-ki/+) and the conditional knockout Smo mutant retinas (Smoflox/flox; Math5cre-ki/+) (data not shown). These results suggested that Smo activity was likely involved at an earlier stage prior to retinal progenitor withdrawal from the cell cycle.
To delete Smo in cycling retinal progenitors, we used a Chx10-Cre transgenic mouse line (Rowan and Cepko, 2004). Under the Chx10 promoter control, this mouse line expresses a fully active Cre-GFP fusion protein at E11.5 with a varying degree of mosaicism (Rowan and Cepko, 2004). At E12.5, when only a few cells expressed the RGC cell marker Brn3a (Liu et al., 2000; Wang et al., 2002), the majority of retinal progenitors were positive for GFP, indicating that Cre expression in retina of the Chx10-Cre mouse was initiated prior to the onset of retinogenesis (Fig. 1A-C). By E14.5, GFP signals persisted at high levels in the ventricular zone occupied by proliferating progenitor cells but diminished greatly in the inner retina occupied by postmitotic neurons (Fig. 1D-G). Therefore, the Cre-GFP fusion protein appeared to be labile in retinal cells that had exited the cell cycle.
We next performed genetic ablation of Smo using the Chx10-Cre mouse to provide Cre in the retina. Quantitative real-time PCRs were performed to evaluate the effects of Smo ablation on expression of Hh signaling components (Fig. 1P). At E15.5, compared to control retinas with two functional copies of Smo alleles (Smoflox/flox; +/+, hereafter referred to as Smo+/+), Smo homozygous conditional knockout mutant retinas (Smoflox/flox; Chx10-Cre, hereafter referred to as Smo−/− cKO mutant) contained a very low level of Smo transcripts (5.0 %) and a severely reduced level of Ptc1 (34.8 %), the Hh receptor and a direct target of Hh signaling (Fig. 1P). Furthermore, expression levels of all three Hh signaling effectors, Gli1, Gli2, and Gli3, showed 69.3 %, 58.6 %, and 16.5 % reduction, respectively (Fig. 1P). These data indicate that Hh signal transduction is severely impaired in Chx10-Cre mediated Smo cKO retinas.
At postnatal day 0 (P0), Smo−/− cKO mutant retinas showed severe phenotypes compared to their heterozygous littermates (Smoflox/+; Chx10-Cre, hereafter referred to as Smo+/−). The Smo−/− cKO retinas were smaller in size and displayed a reduced ventricular zone and an expanded inner retina (Fig. 1H-M). Consistent with the mosaic expression of Cre-GFP, cross sections of Smo cKO retinas showed occasional radially orientated columns without GFP signals (Fig. 1L, M). These Cre-negative regions appeared to resemble the wild type retinas in morphology (Fig. 1L, M, and Suppl. Fig. 1). Furthermore, quantification by flow cytometry using proliferating cell nuclear antigen (PCNA) as a progenitor cell marker showed that at E15.5, control Smo+/+ and Smo+/− retinas contained identical proportions of progenitor cells, whereas Smo−/− cKO retinas contained significantly reduced progenitor cells compared to Smo+/+ controls (60.8 % to 52.5 %) (Fig. 1N). By E17.5, Smo−/− cKO retinas showed a nearly 40 % reduction of GFP-positive cells (from 57.7 % to 34.7 %), the presumptive retinal progenitors (Fig. 1O).
Together, these results provide definitive evidence that during normal mouse retinal development Smo function is required cell autonomously by proliferating retinal progenitors prior to their withdrawal from the cell cycle.
The expanded inner retina in Smo cKO mutants suggested that Hh signaling deficiency affected early retinogenesis. We therefore characterized the effects of Smo ablation on the commitment of progenitor cells towards the RGC fate. Immunostaining by the neuronal marker βTubulin (βTub) detected similar labeling patterns in Smo+/+ and Smo+/− retinas (Fig. 2A, B). In contrast, the Smo−/− cKO retina showed expanded βTub labeling in the inner retina as well as increased βTub-positive processes throughout the ventricular zone (Fig. 2C). Further analyses using the RGC-specific marker Brn3a showed that at E15.5, the RGC layer in Smo−/− cKO mutants was already expanded (Fig. 2D-I). By P0, in contrast to control Smo+/+ retinas, which contained well-defined RGC layer (Fig. 2J, K), Brn3a-positive RGCs in Smo−/− cKO retinas were spread over the inner half of theretina with concomitant shrinking of the proliferative zone (Fig. 2L, M).
Quantitative marker analyses further confirmed the enhanced RGC production in Smo−/− cKOmutants. As early as E15.5, Smo−/− retinas showed significant increases in Brn3a-positive (8.8 % to 13.3 %) and βTub-positive neurons (28.4 % to 33.5 %) (Fig. 2N). At E17.5, heterozygous Smo+/− and control Smo+/+ retinas contained comparable proportions of RGCs as measured by Brn3a and the neurofilament marker NF145 (Fig. 2O). In contrast, loss of both Smo alleles resulted in a near doubling of Brn3a-positive RGCs (7.7 % to 14.8 %) and a significant increase in NF145-positive cells (13.4 % to 19.6 %) compared to controls (Fig. 2O). Consistent with Brn3a immunolabeling patterns at P0, Smo−/− cKO retinas contained a 51% increase of βTub-positive cells (18.5 % to 28.1 %) (Fig. 2P). These results demonstrate that the blockade in Hh signaling prior to the onset of retinogenesis severely affects RGC fate specification among early retinal progenitor cells.
Cone photoreceptor cells are among the earliest born retinal neurons. Previous studies have shown that the homeodomain protein Otx2 regulates transcription of the Crx homeobox gene that is required for photoreceptor differentiation (Nishida et al., 2003; Furukawa et al., 1999; Furukawa et al, 1997a, 1997b; Chen et al., 1997; Viczian et al., 2003). We therefore tested the influence of Hh signaling on cone photoreceptor genesis by examining Otx2 and Crx expression. In situ hybridization of E15.5 Smo+/+, Smo+/−, and Smo−/−cKO mutant retinas revealed similar distributions of Otx2 and Crx transcripts in the ventricular zone and near the ventricular surface, where postmitotic cone cells accumulated (Fig. 3A-F). In addition, immunostaining by an antibody recognizing the cone-specific G protein γ subunit (GγC) also detected similar labeling patterns among Smo+/+, Smo+/−, and Smo−/−cKO mutant retinas at E15.5 (Fig. 3G-I).
In order to detect potential minor effect of Smo deficiency on photoreceptor production, we used flow cytometry to analyze a large number of retinal cells by labeling for photoreceptor precursor marker Crx and cone precursor marker GγC. At both E15.5 and E17.5, mild yet statistically significant increases of Crx-positive cells were detected in Smo−/− cKO retinas compared to control retinas (11.9 % to 14.6 % at E15.5; 18.9 % to 22.7 % at E17.5) (Fig. 3J). At E17.5 Smo−/− cKO retinas showed slightly higher but not statistical significant levels of GγC-positive cells compared to the control retinas (Fig. 3K). These results show that in addition to influencing RGC genesis, early Hh signaling deficiency also results in a mild increase in cone photoreceptor production.
In order to determine whether disruption of Hh signaling in retinal progenitors affects development of all early born retinal neurons, we analyzed the production of horizontal cells and amacrine cells. Quantification by flow cytometry using AP2α, an amacrine cell marker (West-Mays et al., 1999), calbindin, a marker expressed in a subtype of amacrine cells and horizontal cells (Dyer and Cepko, 2001b), and Lim1, an early marker for postmitotic horizontal cells (Poch et al., 2007), did not detect significant alterations at E17.5 (Suppl. Fig. 2A-C). Compared to the controls, immunostaining of Smo−/− cKO retinas at P0 did not reveal changes in AP2α–positive cells despite the dispersion of these cells (Suppl. Fig. 2). Furthermore, immunolabeling for NF145 detected similar patterns of differentiating horizontal cells located within the ventricular zone of Smo+/+, Smo+/−, and Smo−/− cKO retinas (Suppl. Fig. 2), indicating that horizontal cell production was not affected. These results thus demonstrate that Hh signals preferentially regulate the production of a subset of early born retinal neurons but have minimum effects on amacrine cell and horizontal cell specification.
To probe mechanisms underlying the differential influence of Hh signaling on retinal cell type specification, we analyzed expression of bHLH transcription factors. In situ hybridization detected a marked increase of Math5 transcripts, which is required for RGC fate determination (Wang et al., 2001; Brown et al., 2001), in the ventricular zone inSmo−/−cKO mutants at E15.5 (Fig. 4A-C). Furthermore, Math5 transcripts were also detected in the inner retina occupied by postmitotic RGCs, indicating that Math5 expression was sustained in Smo−/− RGCs, which normally only transiently express Math5 (Fig. 4A-C). In contrast, expression patterns of bHLH genes Ngn2 and Math3, which are involved in retinal interneuron development (Inoue et al., 2002), were not significantly altered by Smo deficiency (Fig. 4D-I).
We next used real-time PCR to quantify transcript levels of various bHLH genes expressed in the retina. In Smo−/− cKO retinas Math5 showed the most augmented expression to 1.5 fold of the Smo+/+ control retinas (Fig. 4J). In addition, transcripts of the proneural gene Ngn2 increased by 1.2 fold (Fig. 4J). In contrast, bHLH gene Olig2 and Mash1, which are expressed by progenitor cells (Nakamura et al., 2006; Shibasaki et al., 2007; Ohsawa and Kageyama, 2008),showed 56.2 % and 41.8 % reduction, respectively. Other bHLH genes NeuroD and Math3 also showedmilder yet substantial reductions. Interestingly, expression of the bHLH transcription repressor Hes1 was reduced by 34.8 % in Smo−/− cKO mutant retinas at E15.5, whereas Hes5 was not significantly affected by Smo defects (Fig. 4J). These results indicate that Smo deficiency differentially affects the expression of bHLH transcription factors, especially resulting in significant and sustained upregulation of Math5 expression.
Previous studies have shown that a Pax6 null mutation abolishes the expression of multiple bHLH genes in the retina (Marquardt et al., 2001). Recently Pax6 was shown to positively regulate Math5 transcription (Riesenberg et al., 2009). We therefore examined whether Smo deficiency affected known homeobox genes expressed by retinal progenitors. Real-time PCR detected a 39.0 % reduction of Rx/rax and a 25.9 % decrease of Six3 transcripts, respectively (Fig. 4K). However, total Pax6 expression level in Smo cKO retinas remained similar to the control retinas (Fig. 4K), suggesting that homeobox genes other than Pax6 were more sensitive to Hh signals.
The reduction of the ventricular zone in Smo−/− cKO mutant retinas at P0 (Fig. 1) suggested that Hh signaling played a critical role in controlling embryonic retinal proliferation, even though loss of Smo function did not completely abolish cell division. For example, at E16.5 despite an obviously reduced ventricular zone Smo cKOmutant retinas continued to incorporate BrdU among progenitor cells (Fig. 5A-D). To define specific defects caused by Smo deficiency, we examined the distribution of retinal cells during the cell cycle using flow cytometry-based DNA content analysis. Compared with Smo+/− retinas, total cell populations from Smo−/− cKO mutants showed an increase in G1/G0 and a decrease in S phase cells at E14.5 (Suppl. Fig. 3A). Similar trends were detected at E17.5, when Smo+/+ and Smo+/− cells behaved identically, but Smo−/− mutant cells showed an increased distribution in the G0/G1 phase and concomitant decreases in both S and G2/M phases (Suppl. Fig. 3B).
In order to directly analyze the effect of Hh signaling on proliferating progenitor cells, which express high levels of GFP, we next assayed the cell cycle distribution of GFP-positive cells in Smo−/− cKO mutant and heterozygous Smo+/− retinas. Profiling GFP intensity and DNA contents at E17.5 clearly showed that Smo−/− cKO retinas contained more GFP-negative cells with 2n DNA content, indicating the presence of more postmitotic cells in the G0 phase (Fig. 5E). In contrast, Smo+/− retinas contained more GFP-positive cells with 2n DNA content, representing progenitor cells residing in the G1 phase of the cell cycle (Fig. 5E). Quantification of GFP-positive cells at E14.5 demonstrated that compared with Smo+/− retinas, Smo−/− cKO retinas showed a small but significant increase of G1 phase cells (from 68.2 % to 71.8 %) and a decrease of S phase cells (from 21.9 % to 18.8 %) (Fig. 5F). The cell cycle abnormality became more severe by E17.5; Smo−/− cKO progenitors showed expanded G1 cell population (from 62.6 % to 72.1 %) as well as reduced S (27.3 % to 20.3 %) and G2/M (10.1 % to 8.4 %) cell pools (Fig. 5G). These results demonstrate that Hh signals profoundly affect cell cycle distribution of proliferating progenitor cells. Moreover, the effect of Hh signaling on the cell cycle is greater to late embryonic retinal progenitors.
To investigate whether the abnormal cell cycle distribution of Smo deficient progenitor cells was due to defects in cell cycle progression, we examined expression of cell cycle regulators. By in situ hybridization, we detected very low expression of the Gli1 gene in the ventricular zone of Smo−/− cKO mutant retinas at E15.5, indicating a successful blockade of Hh signaling to retinal progenitors (Fig. 6A, B). Interestingly, the expression level of Shh gene was not upregulated in Smo−/− cKO mutant, despite increased RGCs (Fig. 6E). In situ hybridization revealed that Smo−/− cKO retinas contained fewer cells expressing cyclin D1, a major G1 phase cyclin, in the ventricular zone as well as markedly reduced levels of cyclin D1 transcript in cells still expressing this gene (Fig. 6C, D). This result was confirmed by real-time PCR that detected severe reduction of cyclin D1 transcript by 70.9 % (Fig. 6E). In addition, cyclin E, a late G1 phase cyclin critical for the reentry of S phase, showed more than 53.3 % decrease compared with Smo+/+ controls (Fig. 6E). In addition, cyclin A2, cyclin B1 and cyclin D3 also showed decreased expression (Fig. 6E). Furthermore, expression of the transcription factor E2F1, which is required for the G1 to S transition in a phosphorylation-dependent manner, showed 29.3 % reduction (Fig. 6E). Analyses by flow cytometry also validated that cyclin D1-positive cells were decreased in Smo−/− cKO mutants compared with the Smo+/+ control (47.9 % to 32.7 %) (Fig. 6F). Concomitantly, Smo−/− cKO retinas contained increased number of cyclin-dependent kinase (CDK) inhibitor p27Kip1-positive cells compared to Smo+/+ controls (47.8 % to 63.5 %). In contrast, the number of cells expressing another CDK inhibitor, P57Kip2, did not change (Fig. 6F).
To further define the defective step in cell cycle progression caused by Smo deficiency, we performed BrdU pulse-chase coupled with DNA content analysis, which allowed us to monitor a cohort of progenitor cells as they emerged from the S phase and progressed through the cell cycle. In E16.5 wild type retinas, immediately after a 30 min BrdU labeling, 100% of the BrdU-labeled cells resided in the S phase (Fig. 6G). As the chasing period lengthened, BrdU-positive S phase cells gradually declined, coinciding with the emergence of BrdU-labeled G2/M and G1/G0 phase cells. At 9 hours after the BrdU pulse, about 67% of the BrdU-labeled cells were in G1/G0 phase and about 25% were in G2/M phase. By 18 hours post BrdU labeling, coinciding with the G1/G0 population decline, the BrdU-labeled cells were once again reentering S phase of the cell cycle (Fig. 6G). We performed similar analyses to examine cell cycle progression of Smo−/− mutant cells at 9 hours and 18 hours after BrdU pulse labeling at E17.5. At 9 hours post labeling, we detected no difference in S and G2/M phase distributions between Smo+/− heterozygous and Smo−/− cKO mutant cells, indicating that the S to G2/M transition was largely unaffected (Fig. 6H). However, by 18 hours, among BrdU-labeled cells, more mutant cells accumulated in the G1/G0 phase (60.4 % to 68.1 %), and fewer cells had reentered S-phase (38.1 % to 29.6 %), suggesting a G1/S transition defect (Fig. 6H). Importantly, we also analyzed GFP and BrdU double positive cells, which represented the progenitor cell population excluding postmitotic cells, at 18 hours post BrdU labeling, and obtained similar results (Suppl. Fig. 4). Together, these data indicate that Hh signaling critically affects retinal progenitor cell proliferation by facilitating G1/S phase transition.
To delineate the relationship between increased Math5 expression and enhanced RGC production found in the Smo−/− cKO mutants, we generated Smo and Math5 double mutant retinas. The Smo+/−; Math5+/− double heterozygous retinas showed similar distribution and proportion of RGCs compared to Smo+/+; Math5+/− retinas (data not shown); whereas Smo−/−; Math5+/− retinas contained an increased number of Brn3a-positive RGCs (Fig. 7I, J, Q, T). As previously described in Math5−/− KO mutant retinas, the Smo+/−; Math5−/− retinas showed a dramatic reduction in RGCs compared to Smo+/−; Math5+/− retinas (Fig. 7I, K, Q, T). Quantitative analysis also confirmed that NF145-positive cells were increased in Smo−/−; Math5+/− and decreased in Smo+/−; Math5−/− retinas compared to the double heterozygous Smo+/−; Math5+/− controls (Fig. 7R,T). In Smo−/−; Math5−/− double mutants, the enhanced RGC production observed in Chx10-Cre mediated Smo−/− cKO retinas was completely blocked, and the proportion of RGCs was similar to those found in Smo+/−; Math5−/− retinas (Fig. 7I-L, Q, R, T). These results demonstrate that the proneural bHLH protein Math5 is necessary for the increased RGC genesis found in Smo−/− retinas.
We also analyzed effects of Math5 deficiency on cone cell production in the Smo cKO mutant background. The Smo+/−; Math5+/− double heterozygous retinas showed normally patterns of labeling by the cone cell marker GγC as found in Smo+/+; Math5+/− retinas (data not shown). Consistent with previous Math5 KO results (Brown et al., 1998), Smo+/−; Math5−/− retinas displayed slightly enhanced GγC labeling at the ventricular surface at E17.5 compared to the Smo+/−; Math5+/− double heterozygous retinas (Fig. 7M, O). Quantification further revealed that Smo−/−; Math5+/− and Smo+/−; Math5−/− retinas also contained statistically significant increases of Crx-positive photoreceptor precursors at E17.5, from the Smo+/−; Math5+/− double heterozygous level of 17.3 % to 20.3 % and 24.9 %, respectively (Fig. 7S, T). Interestingly, the Smo−/−; Math5−/− double mutants showed a further enhancement of GγC labeling at the ventricular surface compared to either Smo or Math5 single mutants (Fig. 7M-P). Quantitative analyses confirmed that the Smo−/−; Math5−/− double KO retinas contained 27.4 % Crx-positive cells (Fig. 7S, T), indicating that effects on enhanced cone production due to Smo and Math5 deficiencies were additive.
A previous study has suggested that lost of Math5 function affects the cell cycle exit of early retinal progenitors (Le et al., 2006), we therefore examined whether Math5 played a role in the cell cycle defects detected in Smo mutant retinas. Immunostaining of GFP at E17.5 revealed that in contrast to the Smo−/−; Math5+/− retinas, Smo+/−; Math5−/− and Smo−/−; Math5−/− retinas retained a broad ventricular zone occupied by GFP-positive cells (Fig. 7E-H). However, compared to heterozygous Smo+/−; Math5+/− retinas, Smo−/−; Math5+/− retinas contained half of the GFP-positive progenitor cells (51.9 % to 25.4 %), whereas Smo+/−; Math5−/− retinas showed a lesser but statistically significant reduction of the progenitor pool (51.9 % to 42.5 %)(Fig. 8A, C). The Smo−/−; Math5−/− double KO retinas did not show further loss of GFP-positive progenitors as compared with Smo−/−; Math5+/− retinas (Fig. 8A, C).
To examine the potential effects of Math5 on cell cycle progression, we analyzed the distribution of GFP-positive progenitors in different phases of the cell cycle. As expected, the GFP-positive progenitors in the Smo−/−; Math5+/− retinas consistently showed a significantly higher percentage of G1 cells (from 65.6 % to 76.1 %) and lower percentage of S phase cells (from 29.8 % to 20.3 %) compared to heterozygous Smo+/−; Math5+/− retinas (Fig. 8B, C). However, loss of Math5 in Smo+/−; Math5−/− retinas only resulted in slight decline of S-phase cells (29.8 % to 26.7 %) and no changes in G1 or G2/M distribution (Fig. 8B, C). Moreover, the Smo−/−; Math5−/− double KO retinas showed similar cell cycle distribution as found in the Smo single mutant Smo−/−; Math5+/− retinas (Fig. 8B, C). These results indicate that Math5 function does not impact upon the G1 to S phase transition normally promoted by Hh signaling.
Previous studies of Hh function in vertebrate retinas have largely relied on perturbation of ligands by genetic and non-genetic means, which often result in partial elimination of Hh signals and variable phenotypes. In this study, by ablating the essential Hh signaling component Smo, we have achieved a total blockade of Hh signaling in individual Smo mutant cells. Using the Chx10-cre driver, we show that Hh signaling is required by progenitors in a cell-autonomous manner. Comparing the phenotypes of Smo knockout by Chx10-Cre in cycling progenitors and by Math5-Cre in progenitors exiting the cell cycle, we conclude that Hh signaling is required by progenitor cells prior to their terminal mitosis to generate neurons.
The Chx10-Cre mediated Smo deletion results in severe reduction of progenitors and altered neuronal composition by birth. Therefore, we focused our phenotypic analyses on the embryonic retina to avoid confounding cumulative mutational effects. The most noticeable phenotype is the dramatic increase of RGCs in Chx10-cre mediated Smo cKO retinas. However, despite the expression of Brn3a in the overproduced RGCs (Mu et al., 2008), these neurons may not be fully mature as they have persistent Math5 but abnormally low levels of Shh expression. In addition, photoreceptor precursors showed a mild yet statistically significant increase in Smo cKO retinas. In contrast to altered RGCs and cone cells, we did not detect significant changes in AP2α-positive amacrine cells, Lim1-positive horizontal cells, or calbindin-positive horizontal and amacrine cells. This is consistent with our result that Smo deficiency did not affect p57Kip2, which marks calbindin-positive amacrine cells (Dyer and Cepko, 2001a, b, c). Therefore, in the mouse retina Hh signals profoundly influence the fate determination of a subset of early born neurons, primarily RGCs and cone photoreceptors.
The enhanced RGC genesis in Smo mutant retinas provides compelling genetic evidence that signals derived from postmitotic neurons greatly influence uncommitted progenitors (Zhang and Yang, 2001; Kim et al., 2005; Hashimoto et al., 2006). Among various homeobox and bHLH genes implicated in retinogenesis (Ohsawa and Kageyama, 2008), Hh signaling preferentially suppresses Math5, a key proneural gene required for RGC specification (Brown et al., 2001; Wang et al., 2001). Our results suggest that Hh signals either directly or indirectly regulate Math5 expression. One possibility is that downstream Hh signaling effectors, the Gli proteins, are directly involved in suppressing Math5 expression in Hh responsive progenitor cells. Alternatively, Hh signaling can activate transcription repressor(s) that in turn suppress Math5 transcription. In Smo cKO mutant retinas, expression of the transcription repressor Hes1 is significantly reduced, indicating that Hh signaling positively regulates Hes1, which suppresses proneural genes (Kageyama et al., 2007; Matter-Sadzinski, et al., 2005). Hes1 is a known effector for Notch signaling, which has been shown to inhibit RGC and cone photoreceptor genesis (Austin et al., 1995; Dorsky et al, 1995, 1997; Ahmad et al., 1997; Yaron et al., 2006; Jadhav et al., 2006). Interestingly, we have shown that vascular endothelial growth factor (VEGF), another RGC-secreted factor that promotes progenitor proliferation and suppresses RGC production, also engages Hes1 activity to regulate RGC genesis independent of Notch and ERK (Hashimoto et al., 2006). Thus, a plausible hypothesis is that Hh signals positively stimulate Hes1 expression in progenitor cells, which in turn down-regulates Math5 to suppress the RGC fate (Hashimoto et al., 2006; Fig. 9). Consistent with this hypothesis, a recent study shows that Gli2 may directly promote Hes1 transcription in the postnatal retina (Wall et al, 2009). During cortical neurogenesis, dynamic Hes1 oscillation regulates proneural gene Ngn2 in progenitors (Shimojo et l., 2008). We have detected an increase of Ngn2 in the Smo cKO retinas, suggesting that Hes1 may similarly suppress Ngn2 during retinogenesis. Expression of Math5 normally occurs among embryonic retinal progenitors and is under stringent controls (Hutcheson et al., 2005; Hufnagel et al., 2007; Willardsen et al., 2008), including positive regulation by Pax6 through the 5′ enhancers (Riesenberg et al., 2009). However, the total Pax6 expression level is not affected by Smo deficiency, suggesting that the loss of suppression is responsible for Math5 upregulation. Take together, our results demonstrate that Hh is a major negative regulator of Math5, but the precise mechanism of Hh suppression on this proneural gene requires further analysis at the molecular level.
In addition to giving rise to RGCs, Math5-expressing progenitors also contribute to other retinal cell types, including cones (Yang et al, 2003). In Math5 mutants, abnormal cone photoreceptors have been detected (Brown et al., 2001; Le et al., 2006). Our analyses show that the Math5 single mutant contains a higher proportion of photoreceptor precursors than the Smo single mutant; and that the effects of Smo and Math5 mutations on Crx-positive cells are additive in the double mutants. Thus, Smo and Math5 may independently contribute to cone suppression, possibly by stimulating Hes1 and repressing other bHLH proteins. Moreover, other factors that promote cell cycle exit and the photoreceptor fate may become more available in Smo and Math5 double mutants (Fig. 9). Unlike Math5 and Ngn2, expression levels of several bHLH genes are reduced in the Smo mutant, suggesting that Hh signals normally promote these factors. However, due to the complex relationships among the bHLH proteins, further studies are necessary to delineate how they are influenced by Hh.
Hh signals promote progenitor proliferation in the developing central nervous system (Ruiz I Altaba et al., 2002; Kenney and Rowitch, 2000; Kenney et al., 2003; 2004; Cayuso et al., 2006). However, the role of Hh signals in vertebrate retinal proliferation has remained controversial (Agathocleous et al, 2007). In rodent retinal cultures, Shh-N has a mitogenic effect (Levine et al., 1997; Jensen and Wallace, 1997). But in zebrafish, Shh appears to upregulate p57Kip2 and facilitate cell cycle exit (Shkumatava and Neumann, 2005). In the Xenopus retina, elevated Hh signals appear to accelerate cell cycle progression through both G1/S and G2/M transitions and that fast cycling progenitors have a higher tendency to exit the cell cycle (Locker et al., 2006). In this study, we demonstrate that Smo deficiency causes delayed S phase reentry thus resulting in the accumulation of G1 phase and reduction of S phase cells. Interestingly, both E14.5 and E17.5 progenitors have the same distribution in the G2/M phase (~10%), but a higher proportion of E17.5 progenitors are in the S phase compared to E14.5 (27% versus 23%). The more severe defects found in E17.5 progenitors may reflect the higher rate of S-phase reentry in the late embryonic retina. Our quantitative analyses demonstrate that Smo deficiency causes severe reduction of the G1 phase cyclins, cyclin D1 and cyclin E, which critically control the checkpoint in G1 (Dehay and Kennedy, 2007). We also detected reduced expression of G2/M cyclins, Cyclin A2 and cyclin B1, which is consistent with the observed decrease of G2/M phase cells at E17.5. Results presented here thus provide definitive genetic evidence that Hh signals play an important role in embryonic mouse retinal proliferation.
A fundamental question concerning cell fate determination is whether cell proliferation is intimately linked to cell fate commitment or if they are controlled separately. Some recent studies suggest that cell cycle regulation and cell fate specification can be uncoupled (Godinho et al., 2007;Rompani and Cepko, 2008; Ajioka et al., 2007). One interpretation of Smo cKO mutant phenotypes is that the absence of Hh signaling causes premature cell cycle withdrawal and consequently enhanced neurogenesis. Contrary to this, our findings show that Smo deficiency preferentially affects a subset, instead of all early born cell types, unlike what happens when the cell cycle is completely blocked (Harris and Hartenstein, 1991). Moreover, Hh signals differentially influence the expression of proneural genes, which are involved in specifying distinct retinal cell fates (Ohsawa and Kageyama, 2008). The selective influence of Hh on neuronal fates suggests that separate intracellular machineries are involved in regulating cell cycle progression and cell fate choices.
How might Hh signals coordinate cell proliferation and cell fate selection? Based on current data we favor a model in which Hh signals impact the cell cycle machinery in all progenitors, but critically influence the fate specification only in cells in their neurogenic cell cycle, during which at least one postmitotic neuron is generated. Because Math5 is detectable in a subset of progenitor cells in G2-M phase, it is plausible that Hh signals impact cell fate prior to and/or during G2-M through modulation of Math5 (Fig. 9). A recent study indicates that Math5 also affects cell cycle exit (Le et al., 2006). Our analyses of Smo and Math5 double mutants suggest that Math5 is involved in mediating the effects of Hh on RGC genesis but not on cell cycle progression. Accumulating evidence suggests that bHLH proteins regulate the Cip/Kip family of CDK inhibitors (Guo et al., 1995; Havely et al., 1995; Rothschild et al., 2006; Georgia et al., 2006; Sukhanova et al., 2007; Buttitta and Edgar, 2007). Thus, Math5 is likely to promote cell cycle exit by cooperating with p27Kip1 and specify the RGC fate during the terminal mitosis (Fig. 9).
In summary, our molecular genetic study indicates that Hh signals affect both progenitor cell proliferation and cell fate commitment. Our results support that Hh signals promote cell cycle progression during G1/S transition and regulate specific proneural gene(s) during G2/M towards cell cycle exit. Further investigations of this model will broaden our understanding of how cell extrinsic signals influence neural progenitor cell behaviors to achieve balanced production of diverse neuronal cell types in a given neural network.
Cell-autonomous effect of Smofunction.
Imunofluorescent confocal images of E17.5 retina. (A, C) and (B, D) represent the same sections from Smo heterozygous (+/−)and Smo cKO mutant (−/−) retinas, respectively. (A, B) show cell nuclei staining by DAPI, and (C, D) show merged co-labeling for GFP (green) and Brn3a (red). White arrows (B, D) point to a region that is Cre-GFP negative and therefore expresses functional Smo alleles. This Cre-negative region shows normal ventricular zone and RGC layer, indicating that Smo functions cell autonomously.
gcl, ganglion cell layer; vz, ventricular zone. Scale bar in (A) represents 100 μm for all panels.
Effects of Smo deficiency on amacrine cell and horizontal cell development.
(A-C) Quantification of amacrine and horizontal cell markers by flow cytometry. Percentages of AP2α-, Calbindin-, or Lim1-positve cells among total cells are shown for E17.5 or E16.5 retinas. Genotypes (+/+, Smoflox/flox with no cre; +/−, Smoflox/+ with Chx10-cre; −/−, Smoflox/flox with Chx10-cre) and numbers (n) of individual retinas analyzed are indicated below the bar graphs. No significance sample pairs were detected.
(D-M) Immunofluorescent labeling for amacrine and horizontal cell markers at P0. Images for DAPI (D, G, J), AP2α (E H L), and NF145 (F, I, M) in control (+/+, D-F), Smo heterozygote (+/−, G-I), and Smo cKO mutant (−/−,J-M) retinas are shown.
ac, amacrine cells; gcl, ganglion cell layer; vz, ventricular zone. Scale bar in (D) represents 100 μm for panels (D-O).
Altered cell cycle phase distribution in Smo cKO mutant retins.
Flow cytometric analyses of cell cycle distribution of total retinal cells at E14.5 (A) and E17.5 (B). Cell cycle phase assignments were based on DAPI labeling intensity. Percentages of cells distributed in G1/G0, S, and G2/M phases of the cell cycle among total cells are shown for control (+/+), Smo heterozygous (+/−), and Smo cKO mutant (−/−) retinas. Numbers (n) of individual retinas analyzed are indicated below the bar graphs. **p<0.01, ***p<0.001.
Cell cycle progression defects in Smo cKO mutant retinal progenitors.
Flow cytometric analyses of progenitor cell progression during the cell cycle at E16.5. Smo heterozygous (+/−) and Smo cKO mutant (−/−) retinas were labeled by BrdU for 30 min and followed by 18 hrs of chase period. Percentages of BrdU and GFP double positive cells among GFP-positive cells are shown.
Genotypes (+/+, Smoflox/flox with no cre; +/−, Smoflox/+ with Chx10-cre; −/−, Smoflox/flox with Chx10-cre) and numbers (n) of individual retinas analyzed are indicated below the bar graphs. *p<0.05, **p<0.01.
We thank Drs. Connie Cepko and Andy McMahon for transgenic animals, Dr. Richard Behringer for the Lim1 antibody, Dr. Hiroshi Sasaki for the Gli1 cDNA, Dr. Ingrid Schmid for assistance on flow cytometry, and Dr. Xiang-Mei Zhang for assistance on in situ hybridization. This work was in part supported by grants from Research to Prevent Blindness Foundation, Karl Kirchgessner Foundation, and National Eye Institute to XJY.