|Home | About | Journals | Submit | Contact Us | Français|
Retinal neurons and glia arise from a common progenitor pool in a temporal order, with retinal ganglion cells (RGCs) appearing first, and Müller glia last. The transcription factors Atoh7/Math5 and Ascl1/Mash1 represent divergent bHLH clades, and exhibit distinct spatial and temporal retinal expression patterns, with little overlap during early development. Here, we tested the ability of Ascl1 to change the fate of cells in the Atoh7 lineage when misexpressed from the Atoh7 locus, using an Ascl1-IRES-DsRed2 knock-in allele. In Atoh7Ascl1KI/+ and Atoh7Ascl1KI/Ascl1KI embryos, ectopic Ascl1 delayed cell cycle exit and differentiation, even in cells coexpressing Atoh7. The heterozygous retinas recovered, and eventually produced a normal complement of RGCs, while homozygous substitution of Ascl1 for Atoh7 did not promote postnatal retinal fates precociously, nor rescue Atoh7 mutant phenotypes. However, our analyses revealed two unexpected findings. First, ectopic Ascl1 disrupted cell cycle progression within the marked Atoh7 lineage, but also nonautonomously in other retinal cells. Second, the size of the Atoh7 retinal lineage was unaffected, supporting the idea of a compensatory shift of the non-proliferative cohort to maintain lineage size. Overall, we conclude that Ascl1 acts dominantly to block cell cycle exit, but is incapable of redirecting the fates of early RPCs.
The murine retina is an attractive model for studying cell fate specification. From a common neuroepithelium, four cell types (retinal ganglion cells (RGCs), cone photoreceptors, amacrine and horizontal interneurons) arise largely prenatally. During late embryonic and postnatal development, the bulk of rod photoreceptors, bipolar interneurons and Müller glia are generated (Cepko et al., 1996; Jeon et al., 1998; Sidman, 1961; Young, 1985). When the mouse optic cup forms at E10.5, retinal progenitor cells (RPCs) are multipotent, but over time exhibit restricted developmental potential.
Retinal birthdating experiments demonstrated that neurons and glia exit mitosis in a regular sequence, with one or more cell types arising simultaneously, as well as from sequential cell divisions. Lineage tracing and clonal analyses have also highlighted the multipotency of individual RPCs. Moreover, when early and late RPCs were co-cultured, neither population was temporally reprogrammed, exemplifying the importance of intrinsic determinants in dictating cell fate. Although extrinsic factors influence the relative proportions of cell types produced, they cannot supersede intrinsic regulation of the order in which cell types appear (reviewed in Agathocleous and Harris, 2009; Cayouette et al., 2006; Livesey and Cepko, 2001; Rapaport, 2006).
Numerous retinal studies have focused on basic helix-loop-helix (bHLH) transcription factors, due to their prominent roles in neuronal specification. Indeed, their distinct spatiotemporal expression patterns and loss-of-function phenotypes have solidified key roles during retinal neurogenesis. For example, Atoh7/Math5 appears at the initiation of retinogenesis, and is critically required for RGC formation, and the suppression of cone photoreceptors (Brown et al., 1998; Brown et al., 2001; Kanekar et al., 1997; Kay et al., 2001; Wang et al., 2001). Ascl1/Mash1 expression becomes apparent two days later than Atoh7 in the mouse retina, and is required for normal bipolar interneurons and suppression of Müller glia differentiation (Brzezinski IV et al., 2011; Jasoni and Reh, 1996; Tomita et al., 1996). This suggests that Atoh7 and Ascl1 act via inherently different mechanisms, which is further supported by the evolutionary divergence of their bHLH domains (Bertrand et al., 2002) and their segregated expression: Ascl1 within proliferative RPCs, and Atoh7 in terminally exiting and postmitotic cells (Brzezinski IV et al., 2012; Hufnagel et al., 2010; Jasoni and Reh, 1996; Le et al., 2006; Morrow et al., 1999). Conversely, these factors may be capable of partially or totally substituting for one another, but fail to do so because they are normally segregated into largely non-overlapping lineages (Brzezinski IV et al., 2011). To understand whether these factors might be interchangeable, an in vivo functional substitution is needed, and the consequences assessed during retinal development.
Here we tested whether Ascl1 can reprogram early RPCs to acquire late-born fates, by homologously recombining an Ascl1-IRES-DsRed2 replacement cassette into the Atoh7 locus. The resulting mice displayed ectopic expression of Ascl1, specifically within the Atoh7 lineage, beginning at E11.5 when these cells would normally exit mitosis with the competence to form RGCs. We found that Ascl1 cannot substitute for Atoh7. In Atoh7Ascl1KI/Ascl1KI embryos RGC neurons failed to develop, and the adult mice lacked optic nerves and chiasmata, like Atoh7−/− adults (Brown et al., 2001; Wang et al., 2001). Ectopic Ascl1 failed to produce precocious or excess later-born cell types, although it did induce extra rounds of mitosis, even when coexpressed with Atoh7, in heterozygotes. This overproliferation was temporary, as adult heterozygous eyes contained a normal proportion of RGCs. Intriguingly, our analyses highlighted the ability of ectopic Ascl1 to block cell cycle exit and its inability to instruct RGC genesis in multiple retinal cell lineages.
The Atoh7Ascl1KI targeting vector was created by joining 5 fragments: 1) 2.lKb Atoh7 5’ EcoR1-PstI 5’ arm; 2) Ascl1 cDNA + 3'UTR; 3) IRES2-DsRed2-pA cassette (Clontech); 4) loxP-PGKneo-pA-loxP cassette in opposite orientation; and 5) a 3.1 Kb PvuI-PvuI Atoh7 3’ genomic DNA arm. Figure 1A shows a diagram of the final targeting vector, which was confirmed by complete DNA sequencing. The linearized construct was electroporated into W4 embryonic stem (ES) cells (Auerbach et al., 2000) and colonies selected using G418. To identify homologous recombination at the Atoh7 locus, ES cell genomic DNA was screened by long-range PCR, using one primer outside of each targeting arm and one in DsRed2 coding sequence (Fig 1A, primers C+C’ and D+D’). Both the Atoh7 coding exon and 3'UTR were recombined out of the targeted allele. Southern blotting of ES cell and mouse tail genomic DNA was performed with a 5’ flanking genomic DNA probe (Figs 1A,B). Targeted ES cells were injected into C57BL/6J blastocysts, and chimeric founders mated to C57BL/6J mice. The resulting N2 Atoh7Ascl1KI/+ males were mated with EIIa-Cre female mice (Lakso et al., 1996) to delete the floxed neo cassette, and this was confirmed by PCR genotyping with Cre- or neo-specific primers. Phenotypic analyses were performed by crossing the targeted allele into C57BL/6, 129/Ola and CD-1 genetic backgrounds, then intercrossing Atoh7Ascl1KI/+ mice. No differences in adult optic nerve phenotypes were observed among homozygous mutant N5F2 mice on C57BL/6, 129 or CD-1 genetic backgrounds.
Atoh7LacZ mice were maintained on a CD-1 background, and PCR genotyping performed as described (Brown et al., 2001). For the Atoh7Ascl1KI targeted allele, Forward (5’-AAGGTCTGTTGAATGTCGTGAAGG-3’) and Reverse (5’-TTGAATACGCTTGAGGAGAGCC-3’) primers (Fig 1A,C primers B+B’) were used for 40 cycles at 94°C for 1 min, 56°C for 1 min, and 72°C for 1 min. Atoh7+/+ wild type allele genotyping used Forward (5’-CGCCGCATGCAGGGGCTGAACACG-3’) and Reverse (5’-GATTGAGTTTTCTCCCCTAAGACCC-3’) primers (Fig 1A,C primers A+A’), for 40 cycles at 95°C for 30sec, 60°C for 1 min, and 72°C for 1 min, plus 1X MasterAmp (Epicentre, Madison, WI) in the PCR assay. In some experiments, the Atoh7 retinal lineage was marked using Atoh7-Cre BAC (Brzezinski IV et al., 2012) and Z/EG (Novak et al., 2000), which were crossed into the Atoh7Ascl1KI CD-1 background. The Z/EG reporter expresses EGFP following Cre-mediated recombination.
Embryonic and postnatal tissues were fixed in 4% paraformaldehyde/PBS, processed through a sucrose/PBS series and cryoembedded; or processed postfixation for paraffin embedding and sectioning, with standard histology performed. Immunohistochemistry was performed on 10 micron cryosections as described (Hufnagel et al., 2007). Antibodies used were mouse anti-activated Caspase3 (1:100, Cell Signaling), mouse anti-Ascl1 (1:100, BD Pharmingen, unmasked for 10 minutes at room temperature with 0.5% TritonX/PBS), rabbit anti-βgal (1:5000, Cappel), rat anti-βgal (1:1000, (Saul et al., 2008), rat anti- BrdU (1:100, AbD Serotec, unmasked 1 hour in 1N HCl), sheep anti-Vsx2/Chx10 (1:1000, Exalpha Biologicals), rabbit anti-DsRed (1:500, Invitrogen), chicken anti-GFP (1:1000, Abcam), rabbit anti-Ki67 (1:1000, Vector Labs), rabbit anti-Tubb3 (TuJ1 1:1000, Invitrogen), rabbit anti-M Opsin (1:1000, gift from Cheryl Craft), rabbit anti S-Opsin (1:1000, gift from Cheryl Craft), rabbit anti-Nr2e3 (1:500, gift from Anand Swaroop), rabbit anti-Pax6 (1:1000, Covance), mouse anti-Cdkn1c/p27 (1:100, BD Labs), goat anti-Pou4f2/Brn3b (1:50, Santa Cruz), rabbit anti-Rxrg/RXRγ (1:200, Santa Cruz), mouse anti-Isl1 (1:20; Developmental Systems Hybridoma Bank, Iowa) and rabbit anti-Sox9 (1:200, Chemicon) and DAPI nuclear staining (1:1000, Sigma). Direct-conjugate secondary antibodies (Molecular Probes) or sequential biotinylated secondary (Jackson Immunoresearch) and streptavidin-conjugated Alexa Fluor tertiary antibodies (Molecular Probes) were used to visualize primary antibody labeling.
For all timed matings, E0.5 was assumed to be the date when a vaginal plug was observed. In BrdU pulse- or window-labeling experiments, pregnant dams were injected once with 0.1mg/g body weight of 10mg/mL BrdU in 0.9M NaCl) and were harvested either 1.5 or 24 hours later. For birthdating experiments, pregnant dams or pups were injected with the same concentration of BrdU at a particular age, and then P21 eyes harvested. Images of adult eyes were captured with a Leica dissecting microscope, Optronics digital camera and software.
Microscopy was performed with a Zeiss fluorescent microscope, Zeiss camera and Apotome deconvolution device. For each experiment, we analyzed ≥2 images (200X or 400X) per animal representing both eyes, from ≥3 embryos or postnatal pups from 2 or more independent litters. Images were selected from central retina at equivalent depths based on anatomic landmarks. Cell counts and measurements were performed using the Zeiss Axiovision software (v5.0), using the interactive events and line tools. A paired Student’s T test with Welch’s correction, or ANOVA with a Tukey post hoc test, was performed to calculate P values (Excel or Instat v3.0 software). Photoshop (v7.0) was used to adjust images for brightness and contrast.
To test whether Ascl1 can functionally substitute for Atoh7 during early retinal development, and whether heterochronic misexpression of Ascl1 alters the proliferation or fate choice of early RPCs, we recombined an Ascl1-IRES-DsRed2 cassette into the Atoh7 locus (Figs. 1A-C). The solitary Atoh7 coding exon (Prasov et al., 2010) and Atoh7 3'UTR were precisely replaced, preserving cis regulatory sequences and creating an Atoh7 loss-of-function allele.
In the developing retinas of heterozygous Atoh7Ascl1KI/+ mice, Atoh7+ cells should coexpress Atoh7 and Ascl1. To confirm this, we compared the expression of DsRed2, Ascl1 and E. coli β-galactosidase (βgal) proteins within Atoh7LacZ/Ascl1KI embryonic retinas (Figs 1D,1D’,1F,1F’), taking advantage of an Atoh7LacZ allele to mark Atoh7+ cells (Brown et al., 2001). When neurogenesis begins at E11.5, we observed extensive coexpression of βgal and DsRed2 in dorsocentral optic cup (arrows in Fig 1D’). There was also very widespread coexpression at E12.5, when the initial wave of neurogenesis expanded peripherally (representative arrows in Fig 1F’). At both ages, all βgal+ cells coexpressed DsRed2+, but there were also DsRed2+βgal-negative cells, mostly at the leading edge of neurogenesis (arrowheads in Figs 1D’,1E’). Presumably, retinal cells express the DsRed2 reporter with slightly faster kinetics than βgal, or the anti-DsRed antibody is more sensitive. Importantly, there was complete coexpression of ectopic Ascl1 and DsRed2 (Fig 1E, arrows in 1E’), plus very rare RPCs expressing endogenous Ascl1 protein at E11.5 (arrowheads point to green-only nuclei in Fig 1E’)(Hufnagel et al., 2010; Skowronska-Krawczyk et al., 2009). By E13.5, cells with endogenous Ascl1 expression were more widespread (green only nuclei in Fig 1G). In Atoh7Ascl1KI/+ littermate retinas there was also a cohort of DsRed2+Ascl1+colabeled cells (arrows point to colabeled cells in Fig 1H), in addition to the endogenous Ascl1+ population (arrowheads point to green-only nuclei in Fig 1H). Normally the E13.5 Atoh7 lineage contains few Ascl1+ cells, highlighted by Ascl1 and βgal colabeling of Atoh7LacZ/+ retinal sections (Fig 1I, arrows). This is consistent with a recent demonstration of modest overlap between Atoh7 and Ascl1 lineages in Ascl1-GFP;Atoh7-Cre mice (Brzezinski IV et al., 2011). Overall, we conclude that the Atoh7Ascl1 allele drives ectopic Ascl1 and DsRed2 reporter expression within the Atoh7 lineage, two day earlier than endogenous Ascl1 is observable in the retina.
During retinogenesis, Ascl1 maintains a progenitor pool from which bipolar cells arise, and suppresses Müller glial cell formation (Brzezinski IV et al., 2011; Hatakeyama et al., 2001; Tomita et al., 1996), whereas Atoh7 is critically required for RGC genesis (Brown et al., 2001; Wang et al., 2001). Thus, we tested whether ectopic Ascl1 can alter the fate of Atoh7+ progenitors, or rescue the RGC deficiency in Atoh7 mutants. The adult eyes of Atoh7Ascl1KI/+ mice were grossly similar to wild type Atoh7+/+ littermates, with normal optic nerves and retinal thickness (Fig 2B,2E). There were also no changes in the abundance of RGCs (Pou4f2/Brn3b+, Fig 2H, 2S, Suppl Table 1) and cone photoreceptors (S- and M-Opsin+, Fig 2K,2T, Suppl Table 1). Likewise, there was no shift in Pax6+ amacrine or horizontal cell populations in the inner nuclear layer (INL) (Fig 2U, Suppl Table 1). However, we observed a significant loss of rods (Nr2e3+, Fig 2V, Suppl Table 1) and bipolars (Vsx2/Chx10+, Fig 2N,2W, Suppl Table 1), plus an increase in Müller glia (Sox9+, Fig 2Q,2W, Suppl Table 1) in Atoh7Ascl1KI/+ eyes compared to wild type. To understand these shifts better, we examined their autonomy within the Atoh7-lineage (see below and Fig 3).
The eyes of adult Atoh7Ascl1KI/Ascl1KI mice lack optic nerves (arrow, Fig 2C) but retain the cranial nerves that innervate extraocular muscles (arrowhead, Fig 2C), and thus resemble previously generated Atoh7 mutants (Brown et al., 2001; Wang et al., 2001). Atoh7Ascl1KI/Ascl1KI retinas also had reduced thickness, ectopic vasculature (arrow in Fig 2F), a near total loss of Pou4f2+ RGCs (Fig 2I,2S), and an excess of Rxrg+, S-Opsin+ or M-Opsin+ cones (Fig 2L,2T and data not shown), similar to Atoh7 mutant retinas (Brown et al., 2001; Le et al., 2006). There was no change in INL amacrines (Fig 2U) or Nr2e3+ rods (Fig 2V), but Vsx2+ bipolars and Sox9+ Müller glia were each significantly reduced in Atoh7Ascl1KI/Ascl1KI mice compared to wild type (Fig 2O,2R,2W). As a comparison, rods and bipolars are also reduced in Atoh7−/− retinas (Brown et al., 2001; Brzezinski IV et al., 2005; Brzezinski IV et al., 2012), consistent with prenatal depletion of the RPC pool (Le et al., 2006; Mu et al., 2005). We conclude that Ascl1 cannot substitute for Atoh7 during retinal development.
To determine if the quantitative shifts observed for rods, bipolars and Müller glia in Atoh7Ascl1KI/+eyes, or those for RGCs, cones, bipolars and Müller glia in Atoh7Ascl1KI/Ascl1KI eyes occur cell autonomously, we scored each cell type for coexpression with an Atoh7 lineage reporter. Because Atoh7 is downregulated at birth, the DsRed2 reporter in Atoh7Ascl1KI/+ is uninformative in the adult retina. So we used Atoh7-Cre BAC (Brzezinski IV et al., 2012) and Z/EG (Hadjantonakis et al., 1998) transgenes to permanently label the Atoh7+ descendants in control, Atoh7Ascl1KI/+and Atoh7Ascl1KI/Ascl1KI mice, including cells derived from ectopic Ascl1+ progenitors. The size of the Atoh7 lineage was evaluated by quantifying the percentage of DsRed2+ cells at E12.5 and E13.5 (Fig. 3P, Suppl Table 2), and of Atoh7-Cre;Z/EG (GFP+ cells) at E16.5 and P21 (Fig. 3Q, Suppl Table 2). There was no significant change in the Atoh7 lineage at any age, for either Atoh7Ascl1KI genotype. Thus the increases or decreases for particular retinal cell types (Fig 2S-2W) cannot be explained simply by expansion or contraction of the Atoh7 lineage, which comprises less than 10% of the adult retina (Brzezinski IV et al., 2012).
To examine the cell autonomy of each Atoh7Ascl1KI adult phenotypes, we compared the fractional abundance of histotypic marker+ cells in the GFP+ cohort of Atoh7Ascl1/+ and Atoh7Ascl1KI/Ascl1KI mice also carrying Atoh7-Cre and Z/EG transgenes. As expected, we found a cell-autonomous loss of Pou4f2+ RGCs in Atoh7AscI1KI/Ascl1KI adult eyes (Fig 3A-3C,3R, Suppl Table 1). Next, the pan-cone marker Rxrg/RXRγ(Roberts et al., 2005) was used to assess the autonomy of the cone phenotype in Atoh7Ascl1KI/Ascl1KI eyes. There was no autonomous increase in Rxrg+GFP+ cones (arrows in Figs 3D-3F), although many more Rxrg+GFP– cone photoreceptors (red only) were present in the outer nuclear layer of Atoh7Ascl1KI/Ascl1KI eyes than in controls (Fig 3F). We did note a small autonomous loss of Nr2e3+GFP+ rod photoreceptors in Atoh7Ascl1KI/+ eyes (Figs 3G-3I,3S, Suppl Table 1), which might be a consequence of earlier alterations in cell cycle exit (Fig 4). Finally, we only found rare Vsx2+GFP+; bipolar interneurons and Sox9+GFP+ Müller glia (Fig 3J-3O) in retinal sections from all three genotypes. We conclude that shifts in the total proportion of bipolars and Muller glia (Figs 2O,2R,2W) were a secondary consequence of earlier, developmental defects (see below, Fig 4).
Although Ascl1 misexpression did not autonomously affect the number of bipolars or Müller glia within the Atoh7 lineage, it remained possible that the birthdates of these late classes were accelerated. For this, we birthdated the offspring from Atoh7Ascl1KI/+ intercross matings at E12, E14, E18, P2, P4 and P6, and examined bipolar (Vsx2+) and Muller (Sox9+) cell birth in P21 retinal sections. Rod birthdates were uninformative in this experiment, since these cells are generated over an extended period, beginning at E13 (Cepko et al., 1996; Sidman, 1961; Young, 1985). We found no precocious bipolar neurons (BrdU+Vsx2+) in E12.5 or E14.5 birthdated Atoh7Ascl1KI/Ascl1KI retinas (not shown). There was also no difference among the Atoh7Ascl1KI genotypes for bipolar or Muller glial birthdates from E18.5 to P6 (not shown).
Ascl1 is primarily expressed by dividing RPCs, while Atoh7 is made by RPCs during (< E14) or shortly after (> E15) the terminal cell cycle (Brzezinski IV et al., 2011; Brzezinski IV et al., 2012; Jasoni and Reh, 1996; Le et al., 2006; Pennesi et al., 2003). Therefore it was possible that ectopic Ascl1 could alter the cell cycle status of Atoh7+ cells. We first asked whether the total proportions of mitotic RPCs (BrdU pulse-labeled), exiting RPCs (Cdkn1b+/p27+) and/or nascent RGCs (Pou4f2+) differed among wild type, Atoh7Ascl1KI/+ and Atoh7Ascl1KI/Ascl1KI retinas, from E12.5 to E13.5. Normally at E12.5, the outer proliferative zone (BrdU pulse labeled, green nuclei) is separated from the inner neurogenic zone and nascent RGCs (Pou4f2+, red nuclei), with only sporadic BrdU+Pou4f2+ cells per retinal section (arrows in Figs 4A,4B). In Atoh7Ascl1KI/Ascl1KI eyes, the total proportion of BrdU+ S-phase cells increased at both ages, along with a huge loss of Pou4f2+ RGCs (Figs 4C-E,H). This expansion was accompanied by a loss of both newly postmitotic Cdkn1b+ cells (Fig 4J and data not shown) and Pou4f2+ RGCs (Figs 4C, 4G, 4L). Intriguingly, the reduction in Cdkn1b/p27+ cells directly contrasts an increase observed in Atoh7−/− retinas (Le et al., 2006). For all three makers, the changes observed were sensitive to the dosage of ectopic Ascl1 (Figs 4H,4J,4L, Suppl Table 3).
Next, we tested the cell autonomy of the changes in proliferating, exiting and differentiating cells by quantifying BrdU, Cdkn1b and Pou4f2 coexpression with DsRed2 in E12.5 and E13.5 retinas (Figs 4D-4G, 4I,4K,4M, Suppl Table 3). A cell-autonomous increase in S-phase cells was particularly striking in heterozygous and homozygous mutant cells (arrows in Figs 4D,4E, yellow fractions in 4I), since normally BrdU+ S-phase cells do not express Atoh7 mRNA or Atoh7-GFP transgenes (Hufnagel et al., 2010; Le et al., 2006; Riesenberg et al., 2009a), although a small number of BrdU+βgal+ cells have been found in Atoh7LacZ/+ eyes at this age (Brzezinski IV et al., 2012). Within both heterozygous and homozygous DsRed2-marked lineages, the excess proliferating cells were balanced by a reduction of non S-phase lineage cells (DsRed2+BrdU- red bars in right graph Fig 4I). There was no net change in the size of the Atoh7 lineage, even with two alleles of ectopic Ascl1 (Figs 3P,4I, Suppl Table 2), in stark contrast to the expansion in the total S-phase population (Fig 4H).
The proportions of DsRed2+Cdkn1b+ and DsRed2+Pou4f2+ cells were also examined in adjacent sections (Figures 4J-4M, Suppl Table 3). In both cases, marker-positive cells were autonomously reduced within the Atoh7 lineage, along with a complimentary increase in the marker-negative;DsRed2+ cohort (right graphs in Figs 4K, 4L), further demonstrating there was no fluctuation in the overall size of the Atoh7 lineage. To explain these findings, we also examined the DsRed2-negative cells and saw nonautonomous shifts in the proportions of Cdkn1b+ and Pou4f2+ cells (left graphs in Figs 4K, 4M). This was balanced by a trend for nonautonomous increase in S-phase cells (green bars in left graph Fig 4I). We conclude that ectopic Ascl1 blocks cell cycle exit within the Atoh7 lineage, and that this population is both sensitive and actively responsive, to maintaining a constant proportionality within the developing retina. Simultaneously, we observed nonautonomous changes in the expression of these same markers within cells outside of the Atoh7 lineage. Alternatively, DsRed2 might become activated earlier in the cell cycle, due to an unforeseen issue regarding knock-in allele construction, as well as Ascl1 ectopically promoting RPC proliferation. This could account for maintaining Atoh7 lineage size.
Lastly, we asked whether ectopic Ascl1 caused any early RPCs to stall in the cell cycle, thereby lengthening their overall cycling time. In the developing rodent eye, RPC cell cycle length increases progressively with age (Alexiades and Cepko, 1996). If ectopic Ascl1 induced longer cycle times, it might be interpretable as endowing the early cells with a late RPC characteristic. To test this idea, we pulsed E12.5 retinas with BrdU and scored incorporation 1.5 hours later within proliferative RPCs, which were further identified by Ki67 coexpression (Chenn and Walsh, 2002; Pei et al., 2011). Reductions in the BrdU+Ki67+/Ki67+ proliferating population suggest cell cycle lengthening, but we found no differences among the three genotypes (Suppl Fig 1A). To further examine the balance between cell cycle retention and exit, we performed a 24 hour BrdU pulse-chase and determined the cell cycle retention index (BrdU+Ki67+/BrdU+) (Chenn and Walsh, 2002; Pei et al., 2011). Here too, we found no significant difference among the three Atoh7 genotypes (Suppl Fig 1B). An assessment of the DsRed2-marked population would be the most meaningful, but unfortunately antibody reagent incompatibility prevented us from doing so. Overall, we assume that ectopic Ascl1 does not induce RPCs to lengthen (or shorten) their cell cycle, suggesting that the affected RPCs undergo a defined number of extra cell divisions. However, some or all of these cells might then undergo apoptosis, as a consequence of failing to differentiate correctly.
In Atoh7Ascl1KI/+ eyes, RGC differentiation is initially delayed (Fig. 4H) but eventually recovers, since adult eyes have a normal proportion of these neurons (Fig 2H, 2S). We therefore examined retinal sections at an intermediate time point (E16.5) from Atoh7+/+, Atoh7Ascl1KI/+ and Atoh7Ascl1KI/Ascl1KI embryos carrying Atoh7-Cre and Z/EG transgenes. In wild type and Atoh7Ascl1KI/+ eyes, the Atoh7 lineage (GFP+) contained Pou4f2+ RGCs within the GCL (Figs 5A,5B), indicating that differentiation was occurring in heterozygotes. But Atoh7Ascl1KI/Ascl1KI E16.5 eyes had very few Pou4f2+ RGCs (red nuclei in Fig 5C) or Cdkn1b+ cells (not shown). This difference between the heterozygous and homozygous mutants was further confirmed by examining Isl1 localization in nascent RGCs and displaced amacrines (Elshatory et al., 2007). The Isl1 pattern was normal in heterozygotes, but obviously reduced in homozygotes (Figs 5G-5I). Although the recovery in heterozygotes was predicted, we remained curious about the developmental status of E16.5 homozygous mutant Atoh7 lineage cells.
Intriguingly, E16.5 Atoh7-Cre; Z/EG; Atoh7Ascl1KI/Ascl1KI retinas displayed abnormal GFP+ cell projections into the subretinal (ventricular) space (arrows in 5C; n=3/3 mutants compared to n=0/3 heterozygotes). Colabeling with either DAPI nuclear stain or anti-Tubb3 (βIII Tubulin) demonstrated that these were GFP+ cellular projections, originating from Atoh7 lineage cells displaying neuronal identity (arrows in Fig 5F). This particular phenotype is reminiscent of Pou4f2−/ − eyes, in which profound RGC axon targeting defects indirectly cause excessive RGC cell death (Bermingham et al., 1999; Erkman et al., 2000). In addition, Atoh7−/− mutant retinas were recently shown to accumulate apoptotic cells in the GCL (Feng et al., 2010; Prasov and Glaser, 2012). Therefore, we wondered whether the aberrant axon projections in Atoh7Ascl1KI/Ascl1KI retinas indicated that mutant cells normally fated to become RGCs, instead die. We surveyed E12.5, E14.5 and E16.5 retinal sections from wild type, Atoh7Ascl1KI/+, and Atoh7Ascl1KI/Ascl1KI eyes and found a rapid accumulation of activated-Caspase3+ cells only in Atoh7Ascl1KI/Ascl1KI mutants from E12.5-E16.5 (Fig 5J-5M), which phenocopied Atoh7 mutants (Feng et al., 2010; Prasov et al., 2012).
In this paper we assessed the consequences of genetically replacing the mouse Atoh7 coding region with that of Ascl1 during in vivo retinal development. Our data show that Ascl1 cannot substitute for any aspect of Atoh7 function. Moreover, Ascl1 misexpression within the Atoh7-lineage did not induce late-born retinal fates precociously or prevent RGC formation, although we did find it prolonged the proliferation of early RPCs (Figure 6). Unexpectedly, we found that ectopic Ascl1 blocked cell cycle exit and early differentiation simultaneously within the Atoh7 lineage and nonautonomously in other RPCs.
In Atoh7Ascl1KI;Atoh7lacZ transheterozygous retinas, we observed more DsRed2+ cells than those expressing βgal, although all βgal+ cells coexpress DsRed2. Thus, it appears that DsRed2 overreports the Atoh7 lineage, which could have arisen by several mechanisms: A) differences in knock-in allele construction, B) distinct rates of reporter mRNA and/or protein turnover, C) unequal antibody sensitivity, D) a miRNA in RPCs that normally suppresses Atoh7 expression, thereby allowing DsRed2 to appear earlier, and E) the theoretical possibility that Ascl1 normally suppresses Atoh7 in the Ascl1 lineage, thus ectopic Ascl1 in the Atoh7 lineage dampens the LacZ allele, possibly through an miRNA mechanism. Among these, we favor the possibility that differences in allele construction resulted in mRNAs with unique 3'UTR sequences, since the DsRed2 cassette used SV40 3'UTR, while the LacZ allele retained Atoh7 3’UTR. Furthermore, ectopic Ascl1 utilized endogenous Ascl1 3'UTR sequences, so that the ectopic and endogenous expression closely resembles one another. In the event that there is differential miRNAs regulation of bHLH factor expression, the DsRed2 reporter expression may not completely reflect Atoh7 regulation. There is also a rare possibility of reduced DsRed2 purdurance in reporting the consequences of misexpressing Ascl1 in the Atoh7 lineage. To learn the full extent to which ectopic Ascl1 and endogenous Atoh7 are coexpressed in normal versus gene replacement cells will require double antibody experiments that examine protein expression. Should a specific Atoh7 antisera ever become available (Prasov et al., 2010; Suppl. Fig 2), we will define when and where Atoh7 protein is detectable during the final cell cycle of RPCs, and compare that to endogenous and ectopic Ascl1 expression.
The inability of Ascl1 to rescue RGC development in the absence of Atoh7 is consistent with the low frequency of Atoh7 and Ascl1 coexpression during the earliest phase of retinogenesis (Brown et al., 1998; Brzezinski IV et al., 2011), the absence of RGCs in the endogenous Ascl1 retinal lineage and lack of a RGC phenotype in Ascl1 mutant mice (Brzezinski IV et al., 2011). The population of Ascl1+ cells in the E18.5 retina that derive from the Atoh7 lineage (~25%), presumably produce non-RGC cell types (Brzezinski IV et al., 2011). Thus, Ascl1 has no discernible role in RGC development. Our results demonstrate that Ascl1 also does not prevent ganglion cell development, as this cell class was present in normal abundance in Atoh7Ascl1/+ adult eyes.
One of the best examples of deep evolutionary gene family conservation is the highly conserved role of bHLH factors during retinal neurogenesis. Atoh7 is a competence factor for RGC fate, and a semiortholog of Drosophila atonal (ato), which specifies chordotonal and R8 photoreceptor sensory neurons (Bertrand et al., 2002; Brzezinski IV et al., 2012; Jarman et al., 1993; Jarman et al., 1994). Ascl1 acts during late retinal neurogenesis in the mouse, and is a semi-ortholog of Drosophila scute (sc), which specifies fly external sensory bristles, including interommatidial bristles in the pupal eye (Brown et al., 1991; Campuzano et al., 1985). Previous functional comparisons of ato and sc in Drosophila identified key amino acids within each bHLH domain (Chien et al., 1996; Jarman and Ahmed, 1998; Maung and Jarman, 2007), and specific nucleotides in their E-Box DNA binding sites (Powell et al., 2008; Powell et al., 2004). Together these characteristics endow Ato and Sc with unique abilities during neural specification. Heterochronic misexpression of sc in ato mutant eyes failed to produce R8 neurons, although other photoreceptor cell types were partially rescued (Sun et al., 2000). This partial rescue of non-R8 photoreceptor neurons has been attributed to a subset of downstream target genes, shared by ato and sc during sense organ development (Sun et al., 2000; Sun et al., 2003). In this study however, we found that mouse Ascl1 could not rescue RGC genesis within the Atoh7 lineage.
Direct comparisons of bHLH factor functions during Xenopus retinal development have also been made (Brown et al., 1998; Kanekar et al., 1997; Moore et al., 2002). In the frog eye, neurogenesis occurs rapidly, necessitating simultaneous or successive expression of multiple transcription factors within RPCs. The ato-like factors, Xath5 and XNeuroD promote RGC and amacrine fates respectively, while the sc-like factor Xash1 promotes bipolar neurogenesis (Kanekar et al., 1997; Moore et al., 2002). The neuronal subtype depends on the primary structure, timing of expression and unique posttranslational modifications. For example, the XNeuroD and Xash1 proteins, but not Xath5, are modified by GSK3β- mediated phosphorylation (Moore et al., 2002). While overexpression of either frog Xash1 or mouse Ascl1/Mash1 induced bipolar neurons in frogs, mouse Ascl1 promoted RGC fates when GSK3β activity was blocked (Moore et al., 2002). Moreover, overexpression of mouse Atoh7/Math5 in the frog retina triggered bipolar neurogenesis, not RGC formation (Brown et al., 1998). These data strongly argue that, at least for the frog eye, temporal expression and interactions with other proteins may be more important than the bHLH domain amino acid sequence or E-Box nucleotide sequences per se, in determining neuronal subtype specificity. In the future, it will be important to search for posttranslational mechanisms that regulate bHLH factor activity during mammalian retinogenesis.
The interchangeability of bHLH factors has also been tested to some extent during mouse retinal neurogenesis. Two ato-related bHLH genes, Neurod1 and Atoh3, partially rescued RGC specification when either was homologously recombined into the Atoh7 locus (Mao et al., 2008). The high level of amino acid similarity among Atoh7, Neurod1 and Atoh3 bHLH protein domains, the coexpression of Neurod1 and Atoh7LacZ within a subset of RPCs during normal development (Brzezinski IV et al., 2012; Kiyama et al., 2011), and the cell-autonomous upregulation of Neurod1 in Atoh7 mutants are all consistent with partial functional redundancy for these ato-like bHLH factors (Le et al., 2006; Mao et al., 2008). In a different study, targeted substitution of Ascl1 into the Neurog2 locus (Fode et al., 1998) restored the temporal progression of retinogenesis, which is Neurog2-dependent (Hufnagel et al., 2010). Yet, here we found that Ascl1 cannot rescue the Atoh7 mutant phenotypes.
We propose that mammalian bHLH factors are more likely to substitute for one another, if they are normally expressed in mitotic or nonmitotc RPCs. Whereas Atoh7, Neurod1 and Atoh3 are present in exiting RPCs, and Ascl1 and Neurog2 are found in proliferative RPCs. However, other mechanisms might underlie these differences. For example, Ascl1 and Neurog2 proteins can form complexes (Henke et al., 2009), suggesting that Ascl1 homodimers might substitute for Ascl1-Neurog2 heterodimers in activating downstream target genes. In the developing mouse retina, more work will be needed to fully understand why Ascl1 rescues Neurog2 retinal phenotypes, but not those of Atoh7, since Atoh7 and Neurog2 are each orthologous to ato.
Several studies of retinal regeneration or oncogenesis have shown that high levels of Ascl1 can drive cell cycle exit, others support a pro-proliferative function, including during regeneration. For example, zebrafish retina injury induces the expansion of Müller glia, by coordinated differentiation, division and redifferentiation into neurons. These injuries also activate Ascl1a within Müller glia, prior to cell cycle reentry and neuronal differentiation (Fausett et al., 2008; Yurco and Cameron, 2005). Furthermore, although tumor cell lines often upregulate Notch signaling and downregulate bHLH factors, some neuroendocrine cancer cells exhibit high levels of Ascl1 that correlate with an increased rate of cell division (Jiang et al., 2009; Rapa et al., 2008). In both circumstances, dividing cells with neural identity express Ascl1, implying that it promotes growth within neural-restricted populations. Finally, Ascl1 is normally found in proliferating neuronal progenitors throughout the developing nervous system, where it has been thought to strictly promote cell cycle exit and neural differentiation (Ahmad et al., 1998; Cai et al., 2000; Farah and Easter, 2005; Mao et al., 2009; Tomita et al., 2000). Recent genomic profiling of Ascl1 binding sites, in either developing telencephalon or neural stem cells, indicate that Ascl1 directly regulates a very diverse set of downstream genes, including cell cycle progression, cell cycle exit and oncogenic factors (Castro et al., 2011). Ascl1 may coordinate progression of neuronal progenitors through multiple stages of their development, including proliferation of neural progenitors in the forming retina. Data presented in this study further support a role a pro-proliferative role for Ascl1.
The mechanisms by which Ascl1 acts in the late retina may differ from those discussed here regarding early ectopic expression. When Ascl1 was misexpressed at older ages using retroviruses, rod photoreceptors and bipolar interneurons were generated. Could this be the appropriate developmental window for Ascl1 to promote cell cycle exit? Indeed, Ascl1 mutant retinal explants had fewer late cell types (Hatakeyama et al., 2001; Tomita et al., 1996), and the proliferation defects in Ascl1 mutant cells are manifested late in development (Brzezinski IV et al., 2011). Therefore, it remains unclear whether Ascl1 actively promotes late fates, or simply influences cell cycle exit in this older context. These age-specific differences could be due to integrating low levels of Notch signaling over a long period of time. Alternatively, RPCs may require higher/longer engagement of Notch to prevent differentiation as time proceeds. The failure of Ascl1 to reprogram the early Atoh7 lineage to adopt late fates may reflect an inherent inability to instruct any RPC regardless of age to adopt a late fate, may require additional late environment factors, for example CNTF, LIF or FGF (Ezzeddine et al., 1997; Lillien, 1995; Lillien and Cepko, 1992). or may require cofactors only present in cycling cells.
Perhaps the most surprising aspect of this study is the finding that ectopic Ascl1 can influence cell cycle progression, both autonomously and nonautonomously. How can the Ascl1 transcription factor effect changes in multiple lineages at once, particularly when misexpressed in just one population? Many previous studies have shown that the Notch pathway is influenced by the activity of proneural bHLH factors (reviewed in Pierfelice et al., 2011). In the mouse retina, reduction of Ascl1 activity is accompanied by a downregulation of Dll1 transcription (Nelson et al., 2009; Nelson and Reh, 2008). We propose that ectopic Ascl1 elevates Dll1 expression in the Atoh7 lineage, thereby prolonging proliferation of an adjacent population of RPCs. The stoichiometry of Notch signaling between Atoh7+ cells and RPCs has not been well explored, although Dll1 and Dll4 are present in nascent RGCs (Rocha et al., 2009). It is plausible that Atoh7+ cells normally express a particular level of either ligand on their cell surface, but in Atoh7Ascl1/+ and Atoh7Ascl1KI/Ascl1KI eyes these levels become elevated and drive proliferation in RPCs. An equally intriguing question is whether Notch signaling among Atoh7-lineage cells could explain how this population maintains the correct overall proportionality. Indeed, loss of the Notch integrator Rbpj revealed active homeostasis during photoreceptor cell development (Riesenberg et al., 2009b). Future studies are needed identify and examine molecular mechanisms of Ascl1 and Atoh7 regulation of Dll ligand expression within each endogenous retinal lineage.
A) Quantification of cell cycle length index (% BrdU+Ki67+/total Ki67+) after one pulse of BrdU and a 1.5-hour chase period. B) Quantification of retinal cell cycle retention index (% BrdU+Ki67+/BrdU+) following a single BrdU pulse and 24-hour chase period. In both instances, there was no significant difference among the three genotypes, supporting the idea that ectopic Ascl1 expression does not affect cell cycle length.
We thank Kenny Campbell for W4 ES cells and EIIa-Cre mice; Jane Johnson for Ascl1 antibodies and staining advice; Anand Swaroop for Nr2e3 antibody; Cheryl Craft for Opsin antibodies; the CCHRF Transgenic Core for blastocyst injections; Tien Le and Ashley Riesenberg for invaluable technical support; Chris Chou, Lev Prasov and Kate Maurer for technical advice; and Tom Reh, Lev Prasov, Kenny Campbell and Masato Nakafuku for critical feedback and discussion. This work was supported by a UC Crawley Vision Scholarship to RBH, NRSA F32-EY19227 to JAB, NIH EY14259 grant to TG and NIH EY13612 grant to NLB.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.