We identified Zac1 in a subtractive screen designed to identify regulators of neuronal fate specification 22. In an initial expression survey, we noted high Zac1 expression in the developing retina 25. A detailed spatiotemporal characterization from embryonic day (E) 10.5 through P0 revealed high levels of Zac1 transcripts (Figure 1a–d) and protein (Figure 1f–i) in the outer neuroblast layer (onbl), where proliferating progenitors reside, and not in the inner neuroblast layer (inbl) of postmitotic cells that, prior to P0, primarily includes RGCs and amacrine cells (Additional data file 1 (a–c)). Confirming Zac1 expression in dividing cells, a large number of Zac1⁺ cells incorporated the S-phase label bromodeoxyuridine (BrdU) after a 30 minute pulse at E15.5 (Additional data file 1 (d–f)). Notably, Zac1 expression declined in central, more mature retinal progenitors by P0 (Figure 1d, i).

At P2 (not shown), P7 (Figure 1e, j) and P21 (Additional data file 1 (g,h)), Zac1 transcripts and protein were detected in scattered postmitotic cells in the inner nuclear layer (INL) and RGC layer (GCL; Figure 1k–o). Double immunolabeling with cell type-specific markers at P7 revealed Zac1 expression in CRALBP⁺ Müller glia (64.1% ± 6.26% Zac1⁺cells; n = 3 retinae; Figure 1k,k'), syntaxin⁺ (not shown) and Pax6⁺ amacrine cells (17.5% ± 3.6%; Figure 1m,m'), Brn3a⁺ RGCs (17.2% ± 5.0%; Figure 1o,o') and calbindin⁺ horizontal cells (1.2% ± 0.7%; Figure 1l,l'). Zac1 was not detected in protein kinase C (PKC)-expressing bipolar cells (Figure 1n,n') or in rod and cone photoreceptors in the outer nuclear layer (ONL).

Zac1 is thus expressed biphasically in the retina, initially in dividing retinal progenitors and later in Müller glia, RGCs, amacrine and horizontal cells.

To investigate the in vivo requirement for Zac1, we analyzed embryos with a Zac1 null allele 23. Because Zac1 is maternally imprinted, Zac1+m/- heterozygotes inheriting a wild-type allele from their mother are effectively mutant for Zac1. Indeed, imprinting occurs in the gametes, and complete methylation of Zac1 is achieved in 96.8% of mature oocytes 27. Accordingly, Zac1+m/- retinae were devoid of Zac1 immunolabeling (Additional data file 2) and were thus considered equivalent to null mutants throughout this study.

By P3, 80% of Zac1+m/- pups die, with 50% dying within the first 24 hours 23. To ensure we did not analyze surviving pups with unknown compensatory mechanisms, we studied only Zac1+m/- embryos collected prenatally, when Mendelian ratios of mutants were obtained. To circumvent the problem of retinogenesis not being complete until about P5–6 in the central retina 3, we cultured E18.5 retinae as explants for eight days in vitro (DIV), recapitulating the normal histogenic process 28. Subsequent phenotypic analyses then focused on the central retina, adjacent to the transected optic nerve, where differentiation was complete. Strikingly, most Zac1+m/- explants (55%; n = 27/49) were thicker than their littermate controls, developing a distinct, ectopic cellular layer (ECL) between the INL and GCL (Figure 2b,d,f,f',h,h'; Additional data file 3). Wild-type (Figure 2a,c,e,e',g,g') and remaining Zac1+m/- retinae (not shown) acquired a normal trilaminar structure.

An ECL may develop due to an overall increase in retinal cell number and/or aberrant cellular migration. To determine if Zac1+m/- retinae were hypercellular, DAPI-labeled nuclei were counted. In ECL-containing Zac1+m/- explants (hereafter designated Zac1+m/-+ECL), there was a 1.34-fold increase in the number of INL cells (p < 0.001; 442.9 ± 17.9 cells/field; n = 9 retinae) compared to wild-type controls (329.5 ± 22.0 cells/field; n = 10) or non-ECL containing mutants (henceforth simply designated Zac1+m/-; 314.0 ± 22.1 cells/field; n = 3; Figure 2i). Strikingly, Zac1+m/-+ECL retinae also exhibited a 1.23-fold increase in ONL cells (p < 0.01; 969.4 ± 46.1 cells/field; n = 9) compared to wild-type controls (790.3 ± 40.7 cells/field; n = 10) or Zac1+m/- (735 ± 106.7 cells/field; n = 3; Figure 2i). In contrast, cellular contents of the GCL were comparable in wild-type (59.2 ± 3.1 cells/field; n = 10), Zac1+m/-+ECL (62.0 ± 4.3 cells/field; n = 9) and Zac1+m/- (55.3 ± 1.8 cells/field; n = 3) explants. Zac1 is, therefore, an essential negative regulator of retinal cell number and is also required to orchestrate appropriate cellular migration.

To identify the expanded cell population(s) in Zac1^+m/-+ECL retinae, E18.5→8DIV explants were immunostained with cell type-specific markers. Strikingly, almost all cells in the Zac1^+m/- ECL expressed the homeodomain transcription factor Pax6 (Figure 2f,f'), which was also expressed by amacrine cells in the INL and GCL in wild-type (Figure 2e,e',f,f') and Zac1^+m/- (data not shown) E18.5→8DIV explants. Although Pax6 labels both amacrine cells and RGCs 29, RGCs rapidly undergo apoptosis following optic nerve transection (in explants 30), allowing us to assign an amacrine cell identity to ECL cells. Accordingly, no other RGC markers (Brn3a/3b, Thy1.2; not shown) were detected in the ECL or GCL of wild-type or Zac1^+m/-+ECL explants. Moreover, RGC differentiation is essentially complete by E18.5, and at this stage, equivalent numbers of RGCs were labeled by Brn3a (p = 0.95) and Brn3b (p = 0.23) in wild-type and Zac1^+m/- retinae, indicating Zac1 does not regulate RGC number (n = 3 for each; total n = 6; Additional data file 4). Furthermore, syntaxin, which labels amacrine cell membranes and processes in the inner plexiform layer (IPL; Figure 2g,g'), marked duplicated and disorganized synaptic plexi (IPL/IPL') in Zac1^+m/-+ECL explants (Figure 2h,h'). Finally, amacrine cell subtype markers, including Bhlhb5, calbindin, GABA and the GlyT1 glycine transporter, were all expressed in Zac1^+m/- ECL (Additional data file 3 (k–p)).

Quantitation of Pax6⁺ nuclei in E18.5→8DIV explants revealed a 1.31-fold increase (p < 0.01) in the percentage of amacrine cells in Zac1^+m/-+ECL retinae, while Zac1^+m/-explants contained wild-type proportions of these interneurons (wild type: 15.1 ± 0.5%; n = 4; Zac1^+m/-: 15.2 ± 1.3%; n = 3; Zac1^+m/-+ECL: 19.8 ± 1.3%; n = 3; Figure 2j). In contrast, all other INL cell types were present at equivalent ratios in wild-type and Zac1^+m/-+/-ECL retinae, including bipolar cells (Chx10⁺; wild type: 8.7 ± 0.6%; n = 4; Zac1^+m/-: 10.5 ± 0.9%; n = 3; Zac1^+m/-+ECL: 7.2 ± 1.3%; n = 3), Müller glia (CRALBP⁺: wild type: 5.2 ± 0.3%; n = 4; Zac1^+m/-: 4.3 ± 0.2%; n = 3; Zac1^+m/-+ECL: 3.9 ± 0.5%; n = 3; p27^Kip1+: wild type: 5.9 ± 0.5%; n = 4; Zac1^+m/-: 5.1 ± 0.7%; n = 3; Zac1^+m/-+ECL: 5.8 ± 0.1%; n = 3) and horizontal cells (calbindin⁺; identified also by morphology and apical location; wild type: 1.1 ± 0.3%; n = 3; Zac1^+m/-: 0.8 ± 0.2%; n = 4; Zac1^+m/-+ECL: 0.1 ± 0.1%; n = 3; Figure 2j; Additional data file 3 (m,n)).

Cones normally comprise only 3% of the murine photoreceptor pool 31. In Zac1^+m/-+ECL and wild-type retinae, similar numbers of cones were labeled with peanut agglutinin (PNA; p = 0.26; wild type: 50.4 ± 1.2cells/field; n = 3; Zac1^+m/-+ECL: 64.6 ± 10.1cells/field; n = 3) and s-opsin (p = 0.70; wild type: 44.22 ± 8.91cells/field; n = 3; Zac1^+m/-+ECL: 39.75 ± 6.66cells/field; n = 4; Additional data file 3 (g–j)). Instead, the vast majority of ONL cells in wild-type and Zac1^+m/-+ECL explants expressed the rod-specific markers rhodopsin (Figure 2c, d) and Nr2e3 (not shown), indicating that the rod pool is expanded in Zac1^+m/-+ECL retinae. Zac1 therefore ensures appropriate numbers of rod photoreceptors and amacrine cells are generated during development.

The hypercellularity of Zac1^+m/- retinae could arise due to additional rounds of cell division and/or a reduction in apoptosis. To determine if cell cycle exit was perturbed, S-phase progenitors were BrdU pulse-labeled 30 minutes prior to sacrifice. During embryogenesis (E13.5-E18.5) and in E18.5→2DIV explants, BrdU-labeling indices were similar in wild-type and Zac1+m/- retinae (Figure 3e). In contrast, in E18.5→4DIV Zac1+m/- explants, BrdU incorporation was elevated 2.1-fold over wild type (p < 0.002; Zac1+m/-: 3.4 ± 0.4%; n = 10; wild type: 1.6 ± 0.3%; n = 7; Figure 3a,b,e), although an ECL was not yet distinguishable. Notably, BrdU-labeling indices were variable in individual Zac1+m/- retinae, with about 50% of the mutant explants well above wild-type values (Figure 3f), a phenotypic distribution corresponding well with the proportion of mutant explants that later developed a hypercellular phenotype (see above). Furthermore, in 6DIV explants, when cell division had ceased in wild-type central retinae, BrdU uptake persisted in some mutants (p < 0.05; 0.5 ± 0.01%; n = 4/11; Figure 3c–e). As an independent cell cycle parameter, cyclin D1 (CcnD1) expressing cells were also elevated 1.48-fold (p < 0.001) over wild type (11.7 ± 0.2%; n = 4) in approximately half of the 4DIV Zac1+m/- explants (with phenotype: 17.3 ± 0.6%; n = 4/9; without phenotype: 12.0 ± 0.7%; n = 5/9; Figure 3g–j). Cell proliferation was thus specifically elevated at late stages of retinogenesis in Zac1 mutants.

Ectopic division could occur if progenitors cycled more extensively and/or committed precursors failed to exit the cell cycle. Retinal progenitors are defined by cell cycle-dependent, interkinetic nuclear movements, with G2/M-phase, phospho-histoneH3 (pHH3)-expressing nuclei lining the apical surface (Figure 3k,l), while S-phase nuclei lie more basal in the onbl 32 (Additional data file 1 (e)). This contrasts to committed precursors that migrate towards the vitreal (basal) surface of the inbl to initiate formation of the mature retinal layers. We thus used mitotic position to distinguish proliferating progenitors (apical mitoses) versus precursors (basal mitoses) 33. In Zac1^+m/- retinae, the proportion of pHH3-labeled nuclei was biased towards apical compartments in many Zac1^+m/- 4DIV explants (apical to basal ratio: wild type: 1.02 ± 0.07; n = 10; Zac1^+m/-+phenotype: 2.30 ± 0.35; n = 3/8; Zac1^+m/-: 1.19 ± 0.10; n = 5/8; Figure 3k–e), consistent with an increase in progenitor and not precursor cell divisions. Accordingly, most Pax6⁺ amacrine precursors did not incorporate BrdU after a 30 minute exposure in wild-type or Zac1^+m/- 4DIV explants (Figure 4a,a',b,b'; Additional data file 5 (j)). Similarly, double labeling with Math3, an amacrine and bipolar precursor marker, revealed very few Math3/BrdU double⁺ cells in wild-type and Zac1^+m/- explants (Figure 4c,c',d,d'). Therefore, retinal progenitor cells and not committed precursors are dependent on Zac1 for cell cycle exit.

Compensatory mechanisms exist in the retina to ensure that cellular content remains constant, with excess proliferation often balanced by an increase in apoptosis 3435. Given that Zac1 induces apoptosis when misexpressed in cell lines 24, we tested if it were also required for the normal program of cell death in the retina, using activated-caspase-3 (ac-3), a downstream effector and early marker of commitment to the cell death pathway 36. During embryonic retinal development, apoptosis peaks during the optic cup stage (E10–E11) in the presumptive retinal pigmented epithelium (rpe) and optic stalk and again between E15.5–E17.5, primarily in retinal cells adjacent to the optic nerve head 37383940. We analyzed ac-3 staining in wild-type (n = 6) and Zac1 mutant retinae (n = 6) at E10.5 and E15.5 but did not observe more than a few apoptotic cells per retinal section in either genotype (Additional data file 5 (a-d)). Similarly, at E18.5 (p = 0.14; wild type: 0.4 ± 0.02%; n = 3; Zac1^+m/-: 0.5 ± 0.03%; n = 3) and in E18.5→2DIV explants (p = 0.93; wild type: 3.5 ± 0.4%; n = 3; Zac1^+m/-: 3.5 ± 0.2%; n = 3), comparable levels of apoptosis were observed in both genotypes (Figure 3q). In contrast, after 4 and 8DIV, there were 3.48-fold (p < 0.01; wild type: 4.5 ± 1.0%; n = 7; Zac1^+m/-: 1.3 ± .0.2%; n = 5/6) and 2.02-fold (p < 0.05; wild type: 2.9 ± 0.2%; n = 3; Zac1^+m/-: 1.4 ± 0.3%; n = 4) reductions, respectively, in the number of ac-3⁺ retinal cells in Zac1^+m/- explants (Figure 3o–r).

The reduction in cell death in Zac1^+m/- explants could contribute to the increase in amacrine and rod cell numbers. However, the number of ac-3/Pax6-double⁺ amacrine cells was similar in E18.5→4DIV explants from both genotypes (p = 0.15; wild type: 1.6 ± 0.4%; n = 3; Zac1^+m/-: 0.9 ± 0.1%; n = 3; Additional data file 5 (e–i)). In contrast, there was a 1.82-fold reduction in ac-3⁺ ONL photoreceptors in Zac1^+m/- E18.5→8DIV explants (p < 0.05; wild type: 2.9 ± 0.2%; n = 3; Zac1^+m/-: 1.60 ± 0.4%; n = 4). Zac1 deficiency therefore perturbs pro-apoptotic pathways that adjust cell numbers at late stages of retinogenesis, likely contributing to the increase in rod cell number.

To test if Zac1 was a direct negative regulator of amacrine and rod cell fates, we established a gain-of-function assay, electroporating retinal explants with a pCIG2 vector, containing an internal ribosome entry site (IRES) 2-enhanced green fluorescent protein (EGFP) cassette, or a pCIG2-Zac1 vector, expressing both EGFP and Zac1 (Figure 5). E15.5 and P0 retinal explants misexpressing Zac1 were BrdU-pulse labeled 24 hours post-electroporation, revealing 1.96-fold and 2.49-fold reductions, respectively, in the number of BrdU/EGFP-double⁺ cells compared to controls at E15.5 (p < 0.05; pCIG2: 15.4 ± 1.6%; n = 3; Zac1: 7.9 ± 0.8%; n = 3; Figure 5g) and P0 (p < 0.05; pCIG2: 6.3 ± 0.3%; n = 3; Zac1: 2.5 ± 1.0%; n = 3; Figure 5a–d,g). Zac1 therefore promotes cell cycle exit and/or increases cell cycle length in the murine retina. In contrast, Zac1 misexpression did not increase the number of ac-3⁺ cells compared to controls 24 hour post-electroporation at P0 (p = 0.2; pCIG2: 3.1 ± 0.6%; n = 3; Zac1: 5.2 ± 1.3%; n = 3; Figure 5e,f,h), indicating that Zac1 is not sufficient to induce retinal apoptosis.

To determine if Zac1 was a direct, negative regulator of rod and/or amacrine fates, we examined the molecular phenotype of retinal cells electroporated at P0 and cultured 8DIV. No differences were observed in the ratio of GFP⁺ cells that became Pax6⁺ amacrine cells after electroporation of pCIG2 (p = 0.73; 11.7 ± 3.4%; n = 6; Figure 5i–,k,s) versus pCIG2-Zac1 (10.2 ± 2.3%; n = 6; Figure 5l,s). Similarly, misexpression of Zac1 at E15.5 and E17.5, during the peak of amacrine cell genesis, did not affect amacrine cell number (Additional data file 6). In contrast, Zac1 misexpression at P0 resulted in a 4.49-fold reduction in rhodopsin⁺ rods (p < 0.01; pCIG2: 52.3 ± 4.5%; n = 3; Zac1: 11.6 ± 5.4%; n = 3; Figure 5m,n,s) and a 2.43-fold reduction in Nr2e3-labeled rods (p < 0.05; pCIG2: 30.8 ± 3.6%; n = 3; Zac1: 12.7 ± 3.1%; n = 3; Figure 5s). Zac1-misexpressing progenitors instead preferentially differentiated into Chx10⁺ bipolar cells (1.72-fold increase; p < 0.05; pCIG2: 11.1 ± 1.8%; n = 6; Zac1: 19.1 ± 2.2%; n = 6) and p27^Kip1+ Müller glia (1.95-fold increase; p < 0.05; pCIG2: 14.4 ± 2.2%; n = 6; Zac1: 28.1 ± 4.5%; n = 6), cells types normally generated along with rods postnatally (Figure 5o–s). Zac1 is thus a potent inhibitor of a rod fate but does not directly suppress amacrine cell genesis.

To understand how Zac1 controls amacrine cell numbers, we next determined when ectopic amacrine cells first appeared in Zac1^+m/- retinae. In mouse, amacrine cell genesis normally peaks at E15.5, tapering off before birth 3 (Figure 4j). At E18.5, genes involved in amacrine fate specification/differentiation, including Math3, Foxn4, NeuroD, Pax6 and Barhl2 414243, were expressed in an indistinguishable manner in wild-type and Zac1^+m/- retinae, as were several other genes involved in the specification of all other cell types (Additional data file 7). Cell fate specification was thus grossly normal in E18.5 Zac1^+m/- retinae. In contrast, in E18.5→4DIV Zac1^+m/- explants, Pax6 (Figure 4a,a',b,b',e,f), Six3, Barhl2 and Math3 (Additional data file 8 (i–n)) expression increased, suggesting the amacrine cell population expanded during early postnatal stages in Zac1^+m/- retinae.

To verify that amacrine genesis increased postnatally in Zac1^+m/- retinae, we performed birthdating. E18.5 retinal explants were labeled with BrdU after 1, 2 and 4DIV and then cultivated for 8DIV (Figure 4g,g',h,h'). More BrdU⁺/Pax6⁺ amacrine cells were born at 1DIV (1.76-fold increase; p < 0.05; wild type: 22.7 ± 3.4%; n = 3; Zac1^+m/-+ECL: 39.9 ± 3.8%; n = 4; Zac1^+m/-: 29.8 ± 3.0%; n = 3), 2DIV (2.34-fold increase; p < 0.05; wild type: 8.6 ± 1.3%; n = 7; Zac1^+m/-+ECL: 20.0 ± 4.4%; n = 4; Zac1^+m/-: 10.1 ± 2.8%; n = 6) and 4DIV (5.42-fold increase; p < 0.05; wild type: 1.7 ± 1.0%; n = 6; Zac1^+m/-+ECL: 9.2 ± 2.4%; n = 3; Zac1^+m/-: 4.3 ± 0.4%; n = 2; Figure 4g,g',h,h',i) in Zac1^+m/-+ECL explants compared to wild type, confirming that the period of amacrine cell genesis was prolonged.

Our data suggested that the 'stop' or negative feedback signals that normally limit amacrine cell production later in development 1314 were deficient in Zac1^+m/- retinae (Figure 4j). To thus test if Zac1 was an essential component of the amacrine cell negative feedback loop, we performed aggregation assays. Dissociated E14.5 wild-type retinal cells pre-labeled with BrdU were either cultured alone as intact pellets or in pellet aggregations with a 20-fold excess of dissociated E18.5 wild-type or Zac1^+m/- retinal cells, the latter populations serving as a source of amacrine cell feedback signals (Figure 6a). After 8DIV, pellets were dissociated and Pax6⁺/BrdU⁺ amacrine cells derived from E14.5 progenitors were quantified (Figure 6b–J). Of the E14.5 cells cultured alone, 39.6 ± 3.4% (n = 9; 3 independent experiments) of BrdU⁺cells became Pax6⁺ amacrine cells (Figure 6b–d,k). Consistent with feedback signals being emitted from differentiating, E18.5 wild-type cells, in co-cultures, amacrine cell development from the E14.5-labeled cohort was reduced 1.40-fold (p < 0.01; 27.9 ± 2.0%; n = 10; Figure 6e–g,k). In contrast, amacrine cell development from the E14.5 cohort was restored to normal levels (compared to E14.5 cells alone) in mixed aggregates containing E18.5 Zac1^+m/- cells (p < 0.05, 37.7 ± 2.4%; n = 21), indicative of impaired negative feedback (Figure 6h–k). Zac1 is thus required in postnatal retinal cells to negatively regulate amacrine cell genesis.

The cell non-autonomous requirement for Zac1 as a negative regulator of amacrine cell production implied that this transcription factor must regulate the expression of an unknown secreted signal. We focused on TGFβ cytokines, given their role in feedback control in other systems. Specifically, we studied TGFβII, which regulates cell cycle exit at late stages of rat retinogenesis 44, corresponding to the period when Zac1^+m/- cells proliferated aberrantly. In E18.5 > 4DIV explants, the cognate receptors, TGFβRI and TGFβRII, were expressed at low levels in dividing, Ccnd1⁺ progenitors (Figure 7a,b,d,e) and at higher levels in Pax6⁺ amacrine cells (Figure 7c,f). TGFβII was similarly expressed in Pax6⁺ amacrine cells in the GCL and INL (Figure 7g–i) and at lower levels in Ccnd1⁺ progenitors in the onbl (not shown). Thus, TGFβII signaling could correspond to the amacrine cell stop signal.

In Zac1 mutants, a notable reduction in TGFβII expression was observed in onbl progenitors and in Pax6⁺ amacrine cells in the INL, while GCL levels were similar to wild type (Figure 7j–l). An overall reduction in TGFβII levels was confirmed by western blot, demonstrating that the 25 kDa isoform (note: labile 12 kDa mature form not detected) was reduced in most (n = 8/12) Zac1^+m/- explants (p < 0.05; signal normalized to β-actin; wild type: 1.5 ± 0.04; n = 4; Zac1^+m/-: 0.9 ± 0.2; n = 3/4; Figure 7o,p). To confirm that TGFβ signaling was indeed reduced in Zac1^+m/- retinae, we examined expression of the downstream effector, pSmad 2/3. In E18.5→4 DIV wild-type explants, pSmad2/3 was expressed at diffuse levels throughout the retinae, but at significantly higher levels in the GCL and the basal half of the INL, where differentiated amacrine cells reside, as well as at lower levels in dividing progenitor cells in the onbl (Figure 7m). In contrast, pSmad2/3 levels were decreased in the INL and onbl progenitors in Zac1^+m/- explants (Figure 7n). Accordingly, western blot analysis revealed a significant reduction in pSmad2/3 protein levels in Zac1^+m/- versus wild-type E18.5 > 4 DIV explants when normalized to β-actin (p < 0.01; n = 6/8 mutants analyzed), while total Smad2/3 protein levels were similar in both genotypes (n = 4 for each genotype; Figure 7o,q). TGFβ signaling was thus attenuated in Zac1^+m/-retinae.

To determine if reduced TGFβ signaling results in amacrine cell expansion, conditional TGFβRII mutants were analyzed. Mice heterozygous (Figure 8a,c,e,g) or homozygous (Figure 8b,d,f,h,i,j) for a floxed mutant allele of TGFβRII (hereafter referred to as flTGFβRII; 45) were crossed with mice carrying a R26R reporter and a GLAST::CreERT2 knock-in allele 46. GLAST was expressed in the embryonic retina (Figure 8a, inset) and, accordingly, tamoxifen administered at E16 specifically induced CreERT2 recombinase activity in the E18.5 retina as evidenced by R26R reporter expression (that is, X-Gal staining in tamoxifen injected (Figure 8b) and not un-injected control (Figure 8a) retinae). Accordingly, expression of TGFβRII was reduced in E18.5 flTGFβRII^-/- (Figure 8d) compared to flTGFβRII^+/- retinae (Figure 8c). An overt expansion of the amacrine cell layer, as labeled by Pax6 (Figure 8e,f) and syntaxin (Figure 8g–j), was also evident in tamoxifen-induced E18.5 flTGFβRII^-/- mutant retinae (Figure 8f,h and Figure 8i,j show different mutants) compared to wild-type controls (Figure 8e,g).

While the analysis of TGFβRII mutants supported a role for this signaling pathway in regulating amacrine cell number, we were precluded from analyzing the effects of mutating TGFβRII at postnatal stages as the mutants unexpectedly died at early postnatal stages. We therefore used a complementary pharmacological approach to mimic the late reduction in TGFβ signaling observed in Zac1 mutant retinae. The pharmacological inhibition of TGFβII in the early postnatal rat retina increases proliferation and cell number 44, but specific effects on amacrine cell genesis were not analyzed. In accordance with experiments in rat 44, addition of 0.5 μg/ml soluble TGFβRII-Fc receptor to E18.5→8DIV retinal explants resulted in a 1.55-fold increase in INL/GCL cell number compared to vehicle controls (p < 0.01; control: 387.1 ± 35.0 cells/field; n = 3; TGFβRII-Fc: 601.6 ± 62.1 cells/field; n = 3; Figure 8k,l), while 0.1 μg/ml had no effect (not shown). Moreover, the inhibition of TGFβII signaling resulted in a 1.50-fold increase in the absolute number of amacrine cells (p < 0.01; vehicle control: 220.2 ± 5.9 cells/field; n = 3; TGFβRII-Fc: 331.2 ± 15.3 cells/field; n = 3; Figure 8k–m). These results are consistent with a requirement for TGFβ signaling to negatively regulate amacrine cell number during development.

Next, to show that attenuation of TGFβ signaling underlies amacrine cell expansion in Zac1^+m/- retinae, we performed a rescue experiment. Recombinant TGFβII (or vehicle control) was added to wild-type and Zac1^+m/- E18.5→8DIV explants. In control explants, the percentage of Pax6⁺ amacrine cells was elevated 1.38-fold in Zac1^+m/-+ECL versus wild-type explants (p < 0.01; wild type: 53.6 ± 4.2% INL/GCL cells; n = 4; Zac1^+m/-+ECL; 68.7 ± 4.8% INL/ECL/GCL cells; n = 3; Figure 8p). In contrast, following exposure to TGFβII for 8DIV, the percentage of amacrine cells was equivalent in wild-type and Zac1^+m/-+ECL explants (wild type: 60.4 ± 2.5% INL/GCL cells; n = 3; Zac1^+m/-+ECL: 61.5 ± 2.9% INL/ECL/GCL cells; n = 3; Figure 8n–p). Strikingly, however, an ECL still formed in TGFβII-treated Zac1^+m/- explants (Figure 8o), suggesting that an alternative, non-TGFβ-mediated pathway underlies amacrine cell migration defects. This is also consistent with the inability of TGFβRII-Fc to induce an ECL (Figure 8l). These studies implicate attenuated TGFβII signaling in amacrine cell expansion in Zac1^+m/- retinae.

The widespread expression of Zac1 in dividing progenitor cells in the retina (this study) and throughout the developing neural tube 25535455 suggested that it would have an early role in neural development. Unexpectedly, we found that in the murine retina, Zac1 function is restricted to the early postnatal period. While we cannot rule out the possibility that Zac1 functions redundantly with other factors to regulate early events in retinal development, we would predict that the tumor suppressor-like properties of Zac1 would have to be actively suppressed during early retinal development as most cells that express Zac1 at these stages continue to divide for some time. Indeed, we show here that Zac1 is required to promote cell cycle exit only at late stages of retinogenesis, a context dependency that is also characteristic of other tumor suppressor genes and oncogenes 56. Specifically, we show that, in Zac1 mutants, retinal progenitor cells divide excessively, similar to p27^Kip1 mutants 3552 and in contrast to Rb mutants, where committed precursors instead fail to exit the cell cycle 334748. Our demonstration that cyclin D1 expression is upregulated in Zac1^+m/- retinae provides some insight into the molecular mechanisms underlying Zac1-mediated control of the cell cycle. However, several observations make it unlikely that Zac1 functions directly through p27^Kip1 or the related cyclin dependent kinase (CDK) inhibitor (CDKI) p57^Kip2 to regulate cell cycle exit. Firstly, p27^Kip1 is not required in a temporally restricted manner in the retina, and p57^Kip2 is only required at early stages of retinal development 355257, which contrasts with the late temporal requirement for Zac1. Furthermore, expression of the Kip family CDKIs was not altered in Zac1 mutants, and while there was an increase in p27^Kip1 expression following Zac1 misexpression, it was specific to Müller glia, where this CDKI is normally expressed, and not observed in other cell types. Moreover, a previous cell culture study reported that Zac1 promoted cell cycle exit independently of Kip-family CDKIs or other classic cell cycle regulators such as Rb 24.

The requirement for Zac1 to promote cell cycle exit and apoptosis at late stages of retinal development likely contributes to the formation of hypercellular retinae in mutants, but does not explain why rod photoreceptors and amacrine cells are the only two cell types that are specifically expanded. Strikingly, misexpression of Zac1 robustly inhibited rod differentiation, implicating Zac1 as a bona fide negative regulator of this cell fate. Accordingly, Zac1 expression declines in progenitor cells at P0 when rod photoreceptor genesis begins to peak. Zac1 is also not expressed in differentiated ONL photoreceptors. However, we cannot rule out the possibility that cell non-autonomous mechanisms may also underlie the expansion of the rod pool in Zac1^+m/- retinae. Indeed, we found that the generation of excess rods is directly linked to the formation of an ECL, both occuring in the same approximately 55% of Zac1^+m/- retinae. Notably, we implicated attenuated TGFβ signaling 58, a proapoptotic pathway, in the amacrine cell expansion. However, reduced TGFβ signaling may also underlie the decreased apoptosis we observed in Zac1^+m/- ONLs, consequently contributing to the expansion of the rod pool.

Zac1 misexpression also increased bipolar and Müller glial production in our gain-of-function assays, but rather than proposing that Zac1 is instructive for these fates, we favor the interpretation that progenitor cells prevented from adopting a rod fate instead acquire later-born fates by default. Accordingly, in Zac1^+m/- retinae, we did not observe compensatory decreases in bipolar and Müller glial cell number. Nevertheless, in Xenopus, murine Zac1 also promoted Müller glial as well as RGC genesis, suggesting it might be instructive for a glial identity in different vertebrate species 26. However, there are numerous examples whereby misexpression of a murine gene in Xenopus specifies distinct cell fates compared to misexpression in a mouse model (for example, Mash1 promotes a rod fate when misexpressed in mouse and a bipolar fate in Xenopus 5960. Moreover, in a previous study we showed that murine Zac1 unexpectedly promoted proliferation in Xenopus retina 26, in sharp contrast to its ability to promote cell cycle exit in the murine retina (this study) and cell lines in vitro 2124. To simplify our model of Zac1 retinal function, we therefore consider results obtained in mouse and Xenopus as independent systems where gene function may differ substantively.

Previous studies based on ablation of mature amacrine cells 14 and aggregation of early progenitors with post-mitotic retinal cells 13 demonstrated that amacrine cell number is regulated by negative feedback, but the molecular mechanisms were unknown. Using similar aggregation assays, we showed that Zac1 is required in postnatal retinal cells to limit the number of amacrine cells generated 14. With the exception of rods, numbers of all other retinal cells were not grossly perturbed in Zac1 mutants. The loss of amacrine cell negative feedback therefore does not affect later-born cell types, consistent with previous cell aggregation experiments 113. We thus propose a model whereby initial reductions in amacrine cell genesis, beginning at E16 in wild-type retinae, occurs when progenitors switch to the next competence window to make later-born rods, bipolar cells and Müller glia, an event that is Zac1-independent. This would be followed early postnatally by Zac1/TGFβII-regulated feedback inhibition serving as the final signal to halt amacrine cell genesis (Figure 9).

Feedback pathways exist in diverse biological systems, including the counting factor in Dictyostelium, which dictates group size 2, Drosophila miRNA9a, which regulates sensory organ precursor number by downregulating Senseless expression 61, and the well established role of feedback signals in regulating cell number in vertebrate liver, pancreas, olfactory epithelium and retina 2. Feedback pathways operate by secreting limiting amounts of extrinsic signals that must reach threshold levels to signal cessation of cell genesis 2. Our data support the idea Zac1 acts in post-mitotic amacrine cells during the postnatal period to regulate TGFβII expression, which in turn suppresses amacrine cell genesis. However, our analysis of TGFβRII mutants also indicates that deleting TGFβ signaling earlier in development (that is, from E16 to E18.5), during the peak period of amacrine cell genesis, can also influence amacrine cell genesis. Invoking a threshold model for TGFβII could help explain why defects in cell cycle exit and expansion of the amacrine cell population were not completely penetrant phenotypes in Zac1 mutants. Indeed, developmental processes are known to be highly sensitive to levels of signaling molecules, and stochastic differences in signaling often account for phenotypic variability 62. Moreover, abrogation of the feedback pathway regulating sense organ production in Drosophila, through deletion of miR-9a, similarly results in variable expressivity and penetrance of neuronal overproduction 61.

Notably, amacrine cell migration defects and the subsequent formation of an ECL were independent of attenuated TGFβ signaling in Zac1 mutant retinae. While we attribute the generation of an ECL to the mutation of Zac1, it remains a possibility that ECL formation requires both this genetic deletion as well as the loss of RGCs that occurs in retinal explant cultures, a possibility we cannot directly address given that Zac1 mutants die at birth. Another possibility is that Zac1 directly regulates cell migration by controlling the expression of cell adhesion genes, an idea based on a meta-analysis of microarray data in which several extracellular matrix molecules that could potentially modulate cell adhesion/migration were found to be co-regulated with Zac1 23. The underlying cause of ECL formation is the subject of current investigations.

For embryo staging, the day of the vaginal plug was considered E0.5. Generation of the Zac1 mutant allele was previously described 23. The Zac1 mutant allele was maintained on a C57BL/6 background. Zac1^+m/- heterozygous embryos were generated by crossing Zac1^+/- heterozygous males to C57BL/6 females. Primers for PCR genotyping (35 cycles; 94° for 1 minute, 60° for 1 minute, and 72° for 1 minute) of Zac1 were: wild type 5': AGTGACTCCCCACCTTCTTTCTG; wild type 3': CTTGCCACATTTTTGACAGCG; mutant 5': TGACCGCTTCCTCGTGCTTTAC; mutant 3': CCCCCCAGAATAGAATGACACC. Genotyping of GLAST::CreERT2 and R26R reporter mice were previously described 46. The floxed TGFβRII allele was previously reported 45 and was genotyped by PCR (38 cycles; 95° for 30 s, 62° for 30 s, and 72° for 40 s) with: primer 1 5': TGG GGATAGAGGTAGAAAGACATA-3'; primer 2 5': TATGGACTGGCT TTTGTATTC. To induce deletion of the TGFβRII gene, 3 mg of tamoxifen was administered by oral gavage at E16.0 as previously described 46.

For RNA in situ hybridization, tissue preparation and experimental procedures were followed as previously described 25. Briefly, tissue was fixed in 4% paraformaldehyde (PFA)/1X-phosphate buffered saline (PBS) overnight at 4°C, cryopreserved in 20%sucrose/1X PBS overnight at 4°C and embedded in Cryomatrix™ (Anatomical Pathology USA [Pittsburgh, PA, USA]). Digoxygenin (dig)-labelled probes were generated using a dig-UTP labeling mix and T3, T7 or SP6 RNA polymerases according to the manufacturer's instructions (Roche [Laval, QC, Canada]). Mouse probes included Zac1 25, Hes1, Hes5, Mash1 63, Ngn2 64, Math3 65, Math5 60, NeuroD 66, Pax6 67, Rx 68, Crx 69, Chx10 70. Foxn4 42 and Barhl2 71, Six3 41 and s-opsin 72.

For immunohistochemistry, fixation in 4% PFA/1 × PBS was shortened to 1–2 h at 4°C. Primary antibodies were incubated on slides overnight at 4°C or 1 h at room temperature. The following primary antibodies were used: rabbit active-caspase 3 (1/500; Promega [Madison, WI, USA]), mouse Brn3a (1/500; Chemicon [Temecula, CA, USA]), goat anti-Brn3b (1/250; Santa Cruz [Santa Cruz, CA, USA]), mouse anti-BrdU (5-bromo-2'-deoxyuridine, 1/500; Roche), rat-anti-BrdU (1/10; Oxford Biotech [now Antibodies by Design, Raleigh, NC, USA]), rabbit anti-calbindin (1/1,000; SWANT [Bellinzona, Switzerland]), mouse anti-cyclinD1 (1/100; Santa Cruz), rabbit anti-Chx10 (1/50; Rod McInnes), mouse anti-CRALBP (1/5,000; Jack Saari), rabbit anti-GFP (1/500, Chemicon), goat anti-Math3 (1/100, Santa Cruz), mouse anti-neurofilament 200 (1/500; NF200, Sigma [Oakville, ON, Canada]), rabbit anti-Nr2e3 (1/100; Chemicon), rabbit anti-Pax6 (1/500; Babco [Richmond, CA, USA]]), mouse anti-Pax6 (1/4, Developmental Studies Hybridoma Bank [Iowa City, IA, USA]), rabbit anti-p27^Kip1 (1/500; NeoMarker Lab Vision, [Freemont, CA, USA] ]), mouse anti-protein kinase C (PKC; 1/500; Sigma), mouse anti-rhodopsin (1/500; Chemicon), mouse anti-syntaxin (1/2,000; Sigma), rabbit anti-TGFβII (1/100; Santa Cruz), rabbit anti-TGFβRI (1/100; Santa Cruz), rabbit anti-TGFβRII (1/100; Santa Cruz), rabbit anti-phospho-Smad2/3 (1/100; Santa Cruz), guinea pig anti-GLAST (1/8,000; Chemicon) and rabbit anti-Zac1 (1/1,000 24). Primary antibodies were washed 3 times in PBS with 0.1% triton X-100 (PBT) and detected using secondary antibodies conjugated with Cy3- (1/500; Jackson ImmunoResearch Laboratories, Inc. [West Grove, PA, USA]) or Alexa488 (1/500; Molecular Probes [Invitrogen, Eugene, OR, USA]). Secondary antibodies were diluted in PBT and left on the slides for 1 h prior to 3–10 minute washes with PBT. Note that the TSA™ Tyramide-Fluorescein Immunostaining Kit (NEL701, Perkin-Elmer [Shelton, CT, USA]) was used to amplify anti-TGFβII, TGFβRI, TGFβRII and phospho-Smad2/3 immunostaining as per the manufacturer's instructions. Peanut Agglutinin (PNA) staining was carried out using a 1:200 dilution of the PNA lectin incubated at 37°C for 30 minutes. Sections were then stained for five minutes with DAPI, washed an additional three times with PBS, and mounted with AquaPolymount. β-Galactosidase activity was detected using X-gal as a substrate as previously described 73.

To label S-phase progenitors, pregnant females were injected intraperitoneally with 100 μg/g body weight BrdU (Sigma) 30 minutes prior to sacrifice. For birthdating studies, BrdU was added to the culture media at a final concentration of 10 μM. Embryos were processed for anti-BrdU staining as above except for the addition of a pretreatment with 2N HCl for 30 minutes at 37°C. BrdU immunolabeling after RNA in situ hybridization was carried out using 3,3'-diaminobenzidine (DAB) as a substrate using the Vectastain kit (Vector Laboratories Inc. [Burlingame, CA, USA]).

Retinae were lysed for 15 minutes on ice in RIPA buffer (1% SDS, 1% sodium deoxycholate, 0.1% Nonidet P-40 in 50 mM Tris-HCl (pH 7.6)/150 mM NaCl) plus protease (Complete inhibitor tablet, Roche) and phosphatase (5 mM NaF and 1 mM orthovanadate) inhibitors. Cell lysates were cleared and protein concentrations determined via Bradford analysis. Cell free extract (25 μg) was loaded per lane on a 12% (Smad/phospho-Smad) or 15% (TGFβII) SDS-PAGE gel. Protein was then transferred to PVDF membrane at 80 V for 1 h. Membranes were blocked in 5% skim milk powder or 5% bovine serum albumin (for phospho-Smad) in tris-buffered saline with 0.1% tween 20 (TBST) and then incubated with anti-phospho-Smad2/3 (1/200; Santa Cruz), Smad2/3 (1/200; Santa Cruz), TGFβII (1/200; Santa Cruz) or anti-β-actin (1/5,000, AbCam [Cambridge, MA, USA]) overnight at 4°C. Membranes were washed three times for ten minutes each prior to incubation in horse radish peroxidase (HRP)-conjugated secondary antibodies and development with ECL (Roche).

Retinae were dissected and grown as explants as previously described 74. Briefly, the retinal pigmented epithelium (RPE) and lens were removed from dissected eyes, and the retina was flattened and cultured GCL-up on a Nucleopore Track-Etch membrane (13 mm; Whatman [Maldstone, England]) in explant media (50% MEM, 25% Hanks Solution, 25% horse serum, 6.75 mg/ml glucose, 200 μM L-glutamine, 2.5 mM HEPES) at 37°C in 5% CO₂. The TGFβRII-Fc soluble receptor inhibitor (R&D systems [Burlington, ON, Canada]) was added at 0.5 μg/ml dissolved in PBS (vehicle control) every second day as described 44. Recombinant TGFβII (R&D systems) was added to explants every second day at 1 ng/ml.

Retinae were dissected, dissociated into single cell suspensions and cultured as aggregates essentially as described 1375. Briefly, E14.5 wild-type retinae were dissociated in trypsin (10 min/37°C) and triturated in DMEM/10% fetal calf serum with 100 μl DNAseI (2 mg/ml). Dissociated progenitors were labeled in media with 10 μM BrdU for 1 h. BrdU was washed out and cells were resuspended in culture media at 5 × 10⁵ cells/ml. For co-cultures, 100 μl (5 × 10⁴ cells) of labeled E14.5 progenitors were added to a 20-fold excess (1 × 10⁶ cells) of dissociated E18.5 wild-type or Zac1 mutant cells. Aggregated cells were pelleted by centrifuging for 8 minutes at 2,200 rpm and pellets were transferred after 1 h onto Nucleopore membranes and cultured 8DIV. Pellets were then dissociated and plated on poly-D-lysine-coated slides for immunostaining.

For misexpression, full-length Zac1 cDNA 26 was cloned into a pCIG2 expression vector containing a CMV-enhancer/chicken β-actin promoter and IRES-EGFP cassette (gift from Franck Polleux) 76. For electroporation, eyes were dissected and the RPE removed prior to immersion in 10 μl DNA (3 μg/μl) on a 3% agarose gel plug. Platinum electrodes were placed on either side of the eye (E15.5, 4 mm spacing; and E18.5, 5 mm spacing) and seven 20 ms pulses of 25 V were applied. Electroporated retinae were then cultured as explants.

Immunoreactive cells were counted in sections adjacent to the optic nerve or site of optic nerve transection in explants. In all experiments, cells were counted from a minimum of three embryos (or explants) and three sections per embryo (or explant). The total number of individual retinae analyzed per experiment (n values) is presented in the results section and the total number of cells counted per experiment is presented in the figure legends. All quantification was done from photomicrographs representing a 0.33 mm × 0.25 mm counting field. Statistical variation was determined using the standard error of the mean (SEM). Statistical significance was calculated using a Student's t-test, individually comparing experimental bars against wild-type or control counts.