Zac1 promotes a Müller glial cell fate and interferes with retinal ganglion cell differentiation in Xenopus retina


  • Lin Ma,

    1. Genes and Development Research Group, HBI, IMCH, 2207 HSC, University of Calgary, Calgary, AB, Canada
    Search for more papers by this author
    • L. Ma and J.C. Hocking contributed equally to this study.

  • Jennifer C. Hocking,

    1. Genes and Development Research Group, HBI, IMCH, 2207 HSC, University of Calgary, Calgary, AB, Canada
    Search for more papers by this author
    • L. Ma and J.C. Hocking contributed equally to this study.

  • Carrie L. Hehr,

    1. Genes and Development Research Group, HBI, IMCH, 2207 HSC, University of Calgary, Calgary, AB, Canada
    Search for more papers by this author
  • Carol Schuurmans,

    1. Genes and Development Research Group, HBI, IMCH, 2207 HSC, University of Calgary, Calgary, AB, Canada
    Search for more papers by this author
  • Sarah McFarlane

    Corresponding author
    1. Genes and Development Research Group, HBI, IMCH, 2207 HSC, University of Calgary, Calgary, AB, Canada
    • Genes and Development Research Group, HBI, IMCH, 2207 HSC, University of Calgary, Calgary, AB, Canada, T2N 4N1
    Search for more papers by this author


The timing of cell cycle exit is tightly linked to cell fate specification in the developing retina. Accordingly, several tumor suppressor genes, which are key regulators of cell cycle exit in cancer cells, play critical roles in retinogenesis. Here we investigated the role of Zac1, a tumor suppressor gene encoding a zinc finger transcription factor, in retinal development. Strikingly, in gain-of-function assays in Xenopus, mouse Zac1 promotes proliferation and apoptosis at an intermediate stage of retinogenesis. Zac1 also influences cell fate decisions, preferentially promoting the differentiation of tumor-like clusters of abnormal neuronal cells in the ganglion cell layer, as well as inducing the formation of supernumerary Müller glial cells at the expense of other cell types. Thus Zac1 has the capacity to influence cell cycle exit, and cell fate specification and differentiation decisions by retinal progenitors, suggesting that further functional studies will uncover new insights into how retinogenesis is regulated. Developmental Dynamics 236:192–202, 2007. © 2006 Wiley-Liss, Inc.


In the retina, progenitor cells differentiate into six neuronal populations and one glial cell type, a process that occurs in a stereotyped histogenic order (Young, 1985; Stiemke and Hollyfield, 1995; Cepko et al., 1996). Specifically, retinal ganglion cells (RGCs) are the first to differentiate followed closely by horizontal cells, cone photoreceptors, and amacrine cells, and later by rods, bipolar cells, and Müller glia. As retinal cells differentiate, they migrate to form three distinct cellular layers: a RGC layer (GCL) containing RGCs and displaced amacrine cells, an inner nuclear layer (INL) containing bipolar, amacrine, and horizontal interneurons, as well as Müller glial cell bodies, and an outer nuclear layer (ONL) containing the cell bodies of the rod and cone photoreceptors (Fig. 2A). Each cell layer is separated by a synaptic layer, with the inner plexiform layer (IPL) separating the GCL and INL, while the outer plexiform layer (OPL) separates the INL from the ONL. In the retina, cellular diversity is generated by progressive changes in the competence of an initially multipotent pool of neural progenitors (Competence Model; Turner and Cepko, 1987; Holt et al., 1988; Wetts and Fraser, 1988; Fekete et al., 1994). Thus at any given developmental time, retinal progenitor cells (RPCs) can only give rise to a limited set of postmitotic retinal populations, the identity of which is determined by the repertoire of intrinsic factors that are expressed by RPCs, as well as by extrinsic signals in the environment (reviewed in Cepko et al., 1996; Livesey and Cepko, 2001). Several transcription factors, primarily of the basic helix–loop–helix (bHLH) and homeobox families, are known to be intrinsic determinants of cell fate in the retina (Hatakeyama et al., 2001; Hatakeyama and Kageyama, 2004), but the precise combinatorial codes and differentiation cascades activated downstream of these genes remain incompletely defined.

Figure 2.

Misexpression of mZac1c in Xenopus retina biases progenitors to acquire retinal ganglion cell (RGC) and Müller glial cell fates. A: Schematic illustration of a mature stage 40 Xenopus retina. B: Stage 40 control-injected retina showing green fluorescent protein (GFP) epifluorescence (green) of transfected cells. CF: Stage 40 retina co-injected with expression constructs for GFP and myc-tagged Zac1, showing GFP epifluorescence (green, C,E,F) and immunostaining with anti-myc (red, D,E). F is a high-magnification image of the boxed area in C, showing a representative abnormal cluster of RGC layer (GCL) cells induced by Zac1. G,H: Anti-Zac1 immunostaining (red, G,H) of stage 40 Zac1+GFP (green, H) injected retina. Note that Zac1 antibody does not cross-react with Xenopus Zac1. I: Quantification of GFP+ cells corresponding to each retinal cell class, showing a significant (***P < 0.001) expansion of Müller glial and cells in the RGC layer at the expense of most other cell types after misexpression of mZac1c. n = number of retinae analyzed, and total cell counts were 1,553 for control and 1,146 for mZac1c. Error bars are SEM. gcl, ganglion cell layer; inl, inner nuclear layer; onl, outer nuclear layer; le, lens; PR, photoreceptors; Am, amacrine cells; Bp, bipolar cells; Mu, Müller glia; Hc, horizontal cells; RGC, retinal ganglion cells.

We isolated Zac1 (zinc finger protein that regulatesapoptosis andcell cycle arrest; Spengler et al., 1997) in a subtractive hybridization screen designed to identify genes involved in neural fate specification and/or differentiation in the telencephalon (Mattar et al., 2004). We and others have shown that Zac1, which is also known as pleiomorphic adenoma gene-like 1 (Plag-l1) and lost-on-transformation 1 (Lot1; Abdollahi et al., 1997a; Kas et al., 1998), is expressed at high levels in dividing progenitor cells in the developing nervous system, including the retina (Valente and Auladell, 2001; Ciani et al., 2003; Tsuda et al., 2004; Alam et al., 2005; Valente et al., 2005). Zac1 is an interesting gene in several regards. First, Zac1 is maternally imprinted, a mode of epigenetic control of gene expression levels that is common to many genes involved in growth control (Piras et al., 2000; Smith et al., 2002). In addition, Zac1 has been implicated as a tumor suppressor gene as it is inactivated in several types of carcinomas and can also induce both cell cycle exit and apoptosis when misexpressed in cell lines in vitro (Abdollahi et al., 1997b, 1999, 2003; Spengler et al., 1997; Colitti et al., 1998; Varrault et al., 1998). Finally, Zac1 encodes a transcriptional regulator that belongs to the Krüppel family of zinc finger proteins, which are defined by the presence of seven-C2H2 zinc finger domains. Moreover, in reporter assays, Zac1 has been shown to function as a coactivator or corepressor of other transcriptional regulators, including p53 and several nuclear hormone receptors (Huang and Stallcup, 2000; Huang et al., 2001).

Consistent with a role for Zac1 in development, overexpression in humans, due to a loss of maternal imprinting or duplication of the active paternal allele, is associated with intrauterine growth restriction and transient neonatal diabetes mellitus (Gardner et al., 2000; Kamiya et al., 2000). We hypothesized that Zac1 could also regulate proliferation, apoptosis and/or cell fate decisions in the nervous system based on the mode in which it was isolated (Mattar et al., 2004), its regionalized expression in the neural tube (Valente and Auladell, 2001), and its activity in tumor cells (Abdollahi et al., 1997b, 1999, 2003; Spengler et al., 1997; Colitti et al., 1998; Varrault et al., 1998). Here we provide the first functional evidence that Zac1 can influence a specific developmental process, demonstrating that Zac1 is a key regulator of fate decisions in the developing retina.


Identification and Characterization of Retinal-Specific Zac1 Transcripts in Mouse

We identified murine Zac1 in a subtractive hybridization screen that was designed to identify new genes involved in neural fate specification and/or differentiation (Mattar et al., 2004). Before embarking on an assay to investigate function, we first identified the Zac1 splice variant that was expressed in the developing nervous system of mouse. Previous reports had suggested that human Zac1 has six exons (Abdollahi et al., 2003), while mouse Zac1 has at least three exons (Bilanges et al., 2001). However, the presence of multiple Zac1 transcripts ranging from 3 to 8 kb in mouse tissues (Bilanges et al., 1999), the isolation of numerous murine expressed sequence tags (ESTs), and the identification of alternative splice forms in human (Kas et al., 1998; Bilanges et al., 2001; Abdollahi et al., 2003) and mouse (Huang and Stallcup, 2000), suggested that Zac1 could undergo complex alternative splicing. To identify the splice-form(s) expressed during development, and in the retina in particular, we performed reverse transcriptase-polymerase chain reaction (RT-PCR) on mRNA isolated from embryonic day (E) 18.5 mouse retinae and an E13.5 mouse embryonic cDNA library, using primers surrounding the 5′-ATG and 3′-stop codon. From both templates, an identical 2.2-kb band was amplified (Fig. 1A). Sequencing of the PCR product, which we named mZac1c, revealed an open reading frame contained in three exons that encoded a 673 amino acid protein composed of seven C2H2 zinc fingers in the N-terminus, as well as a linker domain, proline-rich region, and acidic C-terminus, domains that had been defined previously (Huang and Stallcup, 2000; Hoffmann et al., 2003). However, mZac1c differed from the previously reported mZac1a mRNA, which had four coding exons and encoded a 667 amino acid protein (Fig. 1B; GenBank accession no. NM_009538), in that it lacked the last 9 amino acids of mZac1a and instead had an additional 35 amino acids at the C-terminus. An identical C-terminal extension was also reported in mZac1b (Fig. 1B; GenBank accession no. AF147785), but mZac1c lacked an 11 codon PQ-rich sequence insertion found in mZac1b (Huang and Stallcup, 2000). As mZac1c is the predominant transcript in the embryo and developing retina, we investigated its ability to influence retinogenesis.

Figure 1.

Cloning of embryonic mouse Zac1c cDNA. A: Comparison of Zac1 polymerase chain reaction (PCR) products amplified from an embryonic day (E) 13.5 mouse cDNA library (lane a) and cDNA generated from mRNA extracted from E18.5 mouse retinae (lane b). A similar sized PCR product of 2.2 kb was amplified from both templates and designated mouse Zac1c. B: Comparison of the intron–exon structures of the coding regions of mouse Zac1a (NM_009538), Zac1b (AF147785), and our Zac1c transcript, with start and stop codons indicated. The coding region of mouse Zac1c is contained in three exons (black bars) that include previously described exons 7, 8, and 9.

Misexpression of mZac1c Promotes the Genesis of Supernumerary Müller Glia and GCL Clusters in Xenopus Retina

The vertebrate retina is a well-described model for cell fate specification, prompting us to use this system to determine whether mZac1c was sufficient to promote the genesis of a specific neural cell type. We took advantage of a Xenopus gain-of-function assay given that the retinal differentiation period is only 25 hr long in frog (stage 28–stage 40; Holt et al., 1988) vs. approximately 3 weeks in rodents (i.e., E10 to P12; Rapaport et al., 2004). Notably, the order of differentiation of the seven retinal cell types is conserved in these two species. Furthermore, the Xenopus system has been used for primary assessments of the activities of several mouse genes in retinal development, because several, albeit not all, transcriptional regulators have conserved functions across vertebrate phyla (Zuber et al., 2003).

A green fluorescent protein (GFP) reporter was injected either alone (i.e., control; Fig. 2B) or in combination with a myc-tagged mZac1c expression construct (Fig. 2C–H) into 16-cell stage Xenopus embryos, targeting the two dorsal blastomeres fate-mapped to give rise to the retina (Moody, 1987). Note that mZac1c is abbreviated to Zac1 in all figures. Expression of myc-mZac1c was confirmed by means of anti-myc (Fig. 2D,E) and anti-Zac1 (Fig. 2G,H) immunostaining in stage 40 retinae, when the majority of retinal cells were postmitotic. The fate of injected cells was then examined at stage 40, assigning cellular identities based on morphological criteria and laminar position (Holt et al., 1988; McFarlane et al., 1998). In each co-injection experiment, only GFP+ cells that also expressed Zac1, based on immunolabeling with anti-myc or anti-Zac1, were scored. In comparison to misexpression of GFP alone, mZac1c overexpression significantly biased cells toward aggregating in the GCL (84% increase; P < 0.001) or adopting a Müller glial cell fate (311% increase; P < 0.001), cell types that were generated at the expense of photoreceptor (46% decrease; P < 0.001), amacrine (37% decrease; P < 0.001), and bipolar cells (41% decrease; P < 0.001; Fig. 2I).

Remarkably, in injected embryos, most mZac1c-misexpressing Müller glia and cells in the GCL aggregated in large clusters (Fig. 2C), in contrast to control GFP+ cells, which were randomly distributed (Fig. 2B). Moreover, mZac1c-overexpressing cell clusters in the GCL had abnormally large, spread-out cell bodies with multiple short neuritic processes (Fig. 2F), compared with control GFP+ RGCs that had characteristic bipolar morphologies, with 1–2 primary dendrites and one axon extending to the optic nerve (data not shown). Notably, abnormal cells were not found in the GCL in control-injected retinae, suggesting that the differentiation of RGCs was specifically perturbed by mZac1c expression. mZac1c is thus sufficient to influence the final ratio of retinal cell types that are generated, as well as perturb the differentiation of RGCs, suggesting that it is a key regulator of retinal development.

Zinc Finger DNA Binding Domain of Zac1 Is Important for Its Activity in the Retina

The zinc finger domain of Zac1 is important for DNA binding and for heterodimerization (Bilanges et al., 2001). To show that misexpression of mZac1c specifically affected cell fate decisions through its transcriptional activity, we also misexpressed a truncated version of mZac1c lacking the first two zinc fingers, which corresponds to a human-specific isoform designated Zac1Δ2 (Fig. 3A; Bilanges et al., 2001). Deletion of the first two zinc fingers dramatically decreases the ability of Zac1 to dimerize and changes the binding sites for which it has the highest affinity (Hoffmann et al., 2003). In our gain-of-function assay, mZac1cΔ2 had a dramatically diminished capacity to bias retinal cells to acquire a Müller glial or RGC layer phenotype (P < 0.05; Fig. 3B). This finding suggests that the full-length mZac1c protein, containing all seven C2H2 domains, an isoform that can bind DNA with a higher affinity, is better able to bias retinal cell fate choices in the Xenopus retina. These results are consistent with the idea that Zac1 influences cell fate through its transcriptional activity.

Figure 3.

Misexpression of mZac1Δ2 is less efficient at promoting Müller glia and retinal ganglion cell (RGC) differentiation. A: Schematic representation of mZac1c full-length and mZac1Δ2 proteins. B: Quantification of GFP+ cells corresponding to each retinal cell class. Truncation of the first two zinc fingers (Zac1Δ2) severely diminished the ability of mZac1c to influence cell fate specification in the retina (*P < 0.05). n refers to the number of retinae analyzed. Error bars are SEM. PR, photoreceptors; Am, amacrine cells; Bp, bipolar cells; Mu, Müller glia; Hc, horizontal cells; RGC, retinal ganglion cells.

mZac1c Interferes With the Differentiation of RGCs

The abnormal morphology of mZac1c-misexpressing cells in the GCL suggested that they might not have an RGC identity, prompting us to characterize the molecular phenotype of these cells in more detail. We first examined the expression of a panel of pan-neuronal markers in stage 40 Xenopus embryos injected with mZac1c and GFP expression constructs (Fig. 4). In mZac1c-expressing GCL clusters, the cell membranes surrounding the cell somata and processes in the IPL, where RGC dendrites normally extend, were labeled with anti–nerve cell adhesion molecule (NCAM; Fig. 4A–D), anti-neurofilament (NF; Fig. 4E–H), anti–Xen-1 (Fig. 4I–L), and anti-Zn12 (Fig. 4M–P), identifying these cells as neuronal. However, immunolabeling for Zn12, which is a sensory neuronal marker (Hu and Easter, 1999), was reduced in the IPL adjacent to Zac1+ GCL clusters (Fig. 4P), as was Xen-1 immunoreactivity (Fig. 4L). These data suggest that the mZac1c-misexpressing cells in the GCL were neuronal (these cells did not ectopically express glial markers [e.g., Kv4.2, a Müller glia-specific potassium channel; Pollock et al., 2002; e.g., Fig. 5O], but did not send normal dendritic processes into the IPL.

Figure 4.

mZac1c-induced retinal ganglion cell layer (GCL) clusters have a neuronal identity. AP: Stage 40 green fluorescent protein–Zac1 (GFP+Zac1) -injected retina showing GFP epifluorescence (green, A–C,E–G,I–K,M–O) and immunolabeling with the pan-neuronal markers nerve cell adhesion molecule (NCAM, red, B–D), neurofilament (NF, red, F–H), Xen-1 (red, J–L), and Zn12 (red, N–P). C–D, G–H, K–L, O–P are high-magnification images of the boxed areas in B, F, J, N, respectively. Arrowheads in L and P mark disorganized inner plexiform layer. le, lens; gcl, ganglion cell layer; inl, inner nuclear layer; ipl, inner plexiform layer; onl, outer nuclear layer.

Figure 5.

mZac1c perturbs the differentiation of retinal ganglion cells (RGCs) while supernumerary Müller glia are correctly specified. AT: Immunostaining of stage 40 control–green fluorescent protein (GFP, green, A,E,I,M,Q) and Zac1+GFP (green, B–D,F–H,J–L,N–P,R–T) co-injected retinae with the cell-type specific markers Islet (red, A–H), γ-aminobutyric acid (GABA, red, I–L), and Kv4.2 (red, M–T). E and F–H are high-magnification images of the boxed areas in A and D, respectively. Arrows in E show double GFP/islet positive RGCs, while arrows in G show Zac1-misexpressing cells in the RGC layer (GCL) express low levels of Islet. K,L are high-magnification images of the boxed area in J. Arrow in L indicates that a Zac1-induced GCL cluster does not express the amacrine cell marker GABA, while a Zac1-misexpressing cell in the inner nuclear layer (INL) does express GABA (asterisk, L). Q and R–T are high-magnification images of the boxed areas in M and O, respectively. M,N,Q–T: Arrowheads point to Müller glia and their radial processes. Note the GFP+ Müller glial cell clusters induced by Zac1 misexpression in N–O (arrowheads). Red label in N is myc immunoreactivity, indicating Zac1-misexpressing cells. Arrowheads in T show that Zac1-induced Müller glial cell clusters express Kv4.2. le, lens; gcl, ganglion cell layer; inl, inner nuclear layer; onl, outer nuclear layer.

We next examined whether mZac1c-misexpressing GCL clusters contained neuronal cells with a RGC phenotype. The vast majority of control transgene-expressing cells in the GCL that expressed GFP at moderate (89.9%, n = 79) or elevated (88.2%, n = 68) levels were positive for the RGC marker Islet1, a homeodomain transcription factor (Fig. 5A,E; Dorsky et al., 1997). On the contrary, in mZac1c-injected embryos, only 10.6% (n = 123) of intensely GFP-positive GCL cells (i.e., expressing the highest ectopic levels of Zac1), expressed Islet1 at high levels, and only 27.1% (n = 166) of cells that expressed GFP at moderate levels co-expressed Islet1 at appreciable levels (Fig. 5B–D,F–H). However, low Islet1 expression, just above background levels, was detected in many mZac1c-misexpressing RGCs at stage 40 (Fig. 5G, arrows). Islet1 expression levels did not increase in mZac1c-misexpressing GCL cells by stage 42 (data not shown), suggesting that mZac1c did not simply delay the acquisition of an RGC fate, but rather prevented the full differentiation of these cells. Consistent with this interpretation, other RGC markers, such as the POU domain transcription factor Brn3b, which was expressed by 49.2% (n = 333) and 45.8% (n = 528) of strongly positive and moderately positive GFP control cells, respectively, was only expressed in 5.6% (n = 162) of the intensely GFP+ cells, and 15.2% (n = 282) of the moderately expressing GFP+ cells in mZac1c-overexpressing embryos. Consistent with the idea that the ectopic cells in the GCL had not acquired other retinal cell identities, photoreceptor markers (data not shown) and the amacrine cell marker γ-aminobutyric acid (GABA) were not detected in mZac1c-induced GCL clusters (Fig. 5I–L). We thus favor the interpretation that mZac1c-overexpressing cells in the GCL acquire a partial RGC fate (as revealed by low Islet1 expression) and that, while mZac1c perturbs the differentiation of RGCs, it does not influence the neuronal vs. glial fate decision by committed RGC precursors.

Ectopic Müller Glial Cells Differentiate Normally Following mZac1c Misexpression

In control injections, few GFP-expressing cells differentiated into Kv4.2-expressing Müller glia in stage 40 retinae (Fig. 5M,Q). In contrast, after misexpression of mZac1c and GFP, a large number of Müller glia were generated, characterized by typical bipolar morphologies, cell bodies correctly positioned in the INL and highly branched endfeet at the vitreal and ventricular surfaces (Fig. 5N–O,R). In addition, the mZac1c-misexpressing Müller glia identified by morphology were immunopositive for Kv4.2 (Pollock et al., 2002), labeling that was particularly evident in the Müller glial processes and endfeet (Fig. 5R–T). Indeed, scoring of Müller glial cells based on Kv4.2 expression revealed a significant increase in mZac1c vs. control injected retinae that was similar to values obtained using morphological criteria alone (4.17% ± 0.31% in control retinae (n = 6) vs. 20.27% ± 2.53% in mZac1c-misexpressing retinae (n = 5; P < 0.001)). Of note, mZac1c-induced clusters in the INL for the most part did not express neuronal markers, consistent with the glial identity of these cells (data not shown). Thus in contrast to its deleterious effects on RGC differentiation, mZac1c promotes the differentiation of Müller glia with appropriate molecular phenotypes.

Forced Expression of mZac1c Blocks Cell Cycle Exit at an Intermediate Stage of Retinogenesis

We speculated that mZac1c-misexpressing Müller glia and GCL cells formed aggregates due in part to a protracted period of proliferation by cells that should have exited the cell cycle. To examine this possibility, we labeled dividing cells in S-phase of the cell cycle by exposing injected Xenopus embryos to a short pulse of bromodeoxyuridine (BrdU). Embryos were harvested at stages 28, 32, and 40; immunolabeled with anti-BrdU; and a BrdU-labeling index (i.e., number of BrdU+ GFP+ nuclei/total GFP-positive cells) was calculated. At stage 28, when more than half of the retinal progenitor pool was in S-phase of the cell cycle, no significant difference in the BrdU-labeling index was observed in control (62.5% ± 2.6%) vs. mZac1c-overexpressing (55.8% ± 4.3%) retinae (Fig. 6A,D,I). In contrast, at stage 32, the percentage of mZac1c-overexpressing progenitors in the cell cycle (51.9% ± 2.8%) was significantly higher (P < 0.01) than in control-injected embryos (38.8% ± 3.1%; Fig. 6B,E,I). The increase was transient, however, as the number of dividing cells was not significantly different in control (1.01% ± 0.5%) and mZac1c-misexpressing (1.6% ± 0.8%) retinae at stage 40, when the retina was almost fully mature and proliferation in the central retina had for the most part ceased (Fig. 6C,F,I). Thus mZac1c selectively biases retinal progenitors to undergo extra cell divisions during a defined, intermediate period of retinogenesis.

Figure 6.

mZac1c promotes the proliferation of retinal progenitors at an intermediate stage of retinogenesis. AH: Embryos injected with green fluorescent protein (GFP, green, A–C) or co-injected with Zac1+GFP (green, D–F) were exposed to a short bromodeoxyuridine (BrdU) pulse before harvesting at stage 28 (A,D), stage 32 (B,E,G,H) and stage 40 (C,F). Immunolabeling of BrdU is in red (all panels). Inserts in A–F show high-magnification images of GFP/BrdU double positive cells. Arrowheads shows BrdU (red)/GFP (green) double-positive cells in the inner nuclear layer (F) and in a retinal ganglion cell layer (GCL) cluster (the boxed area in G is shown in high magnification in H). I: Quantification of BrdU-labeling index for control and Zac1-misexpressing embryos at each developmental stage. n refers to the number of retinae examined, and total cells counted were as follows: st28: control, 507 and Zac1, 715; st 32: control, 963 and Zac1, 1,280; st 40: control, 1,524 and Zac1, 1,301. Error bars are SEM. **P < 0.01. gcl, ganglion cell layer; inl, inner nuclear layer; onl, outer nuclear layer; vz, ventricular zone.

To determine whether mZac1c-misexpressing GCL and Müller glial cells formed aggregates because precursors that should differentiate instead underwent additional rounds of cell division, we specifically quantified the BrdU-labeling index in mZac1c-misexpressing cell clusters in the INL and GCL. At stage 40, the majority of control cells with bipolar processes spanning the retina had exited the cell cycle (i.e., postmitotic Müller glia; BrdU labeling index = 0.43 ± 0.43%), whereas the BrdU-labeling index for mZac1c-misexpressing cells in the INL was substantively higher (11.0% ± 6.2%), and many clusters composed of morphologically identifiable Müller glia contained at least one proliferating cell (Fig. 6F, arrowhead). Moreover, at stage 33/34, 64% (n = 39) of control GFP-positive cells with processes that spanned the retina failed to incorporate BrdU (i.e., Müller glia identity as opposed to neuroepithelial precursor), whereas significantly fewer (i.e., 42%; n = 67) mZac1c-misexpressing cells spanning the retina were BrdU-negative, indicating that mZac1c delays cell cycle exit and subsequent Müller glial differentiation.

A similar increase in cell proliferation was seen in the GCL at stage 32 after mZac1c misexpression, with considerably more GFP-positive cells in the GCL incorporating BrdU (18.7% ± 4.8%, in control retinae [n = 6] vs. 43.2% ± 3.4%, in mZac1c-misexpressing retinae [n = 6]; P < 0.01; Fig. 6G,H). In contrast, the BrdU-labeling index was similar in control and mZac1c-misexpressing GCL cells at stage 40 (0.45% ± 0.4%, n = 22 in control vs. 0.75% ± 0.8%, n = 19 in mZac1c-misexpressing retinae), indicating that the effect of this transcription factor on cell cycle exit was transient. Misexpression of mZac1c thus inhibited cell cycle exit in the GCL and INL, resulting in the generation of large, possibly clonally expanded cell clusters.

mZac1c Promotes Apoptosis in Xenopus Retina

The expansion of GCL and Müller glial cell types at the expense of amacrine, bipolar, and photoreceptor cells could occur due to the selective cell death of a particular population of retinal cells. Indeed, when misexpressed in cell lines, Zac1 is one of the few tumor suppressor genes that can promote both cell cycle exit and apoptosis (Spengler et al., 1997). We, therefore, investigated whether mZac1c was sufficient to induce apoptosis when misexpressed in Xenopus retina, focusing on stage 32, a time when mZac1c promoted cell proliferation. At stage 32, in control injected retinae (n = 7), 2.33 ± 0.69% of GFP-positive cells coexpressed activated caspase-3 (ac-3), a downstream effector and early marker of commitment to apoptosis (Fig. 7A–C,J; Watanabe et al., 2002). In contrast, in mZac1c and GFP co-injected retinae, 8.67 ± 2.68% (n = 7; P < 0.05) of GFP-positive cells expressed ac-3 (Fig. 7D–F,J), consistent with the idea that mZac1c promotes apoptosis in the retina as previously shown in cell lines in vitro. However, similar to its stage-specific effects on proliferation, by stage 40, the vast majority of cells expressing high levels of mZac1c, including clusters in the GCL, were not immunolabeled by anti–ac-3 (Fig. 7G–I). Thus mZac1c has an enhanced ability to promote apoptosis in the retina, but only at an intermediate stage of retinogenesis.

Figure 7.

mZac1c promotes apoptosis in Xenopus retina at stage 32. AF: Stage 32 control green fluorescent protein (GFP, A–C) and Zac1+GFP–injected (D–F) retinae labeled with anti–activated caspase-3 (ac-3, red, A,C,D,F) and GFP epifluorescence (green, B,C,E,F). The insert in F shows a high-magnification image, and arrowheads in F show GFP/ac-3 double positive cells. GI: Stage 40 Zac1+GFP–injected retinae labeled with ac-3 (red, G,I) and GFP epifluorescence (green, H, I). J: Quantification of apoptotic control and Zac1-misexpressing cells in stage 32 retinae. n refers to the number of retinae examined, and total cells counted were control (n = 2,115) and Zac1 (n = 1,705). Error bars are SEM. *P < 0.05. gcl, ganglion cell layer; inl, inner nuclear layer; onl, outer nuclear layer; vz, ventricular zone.


Here we demonstrated for the first time that the Zac1 tumor suppressor gene regulates a normal developmental process, controlling proliferation, apoptosis, cell fate specification, and differentiation in the retina. Strikingly, Zac1 promoted rather than inhibited the division of retinal progenitors, contrary to its reported role as a tumor suppressor gene in cancer cells and its activity in cell lines in culture. Zac1 also influenced retinal cell fate choices, promoting the genesis of abnormally differentiated neurons in the GCL as well as correctly specified Müller glia. Zac1 function thus changes in a dynamic manner during retinal lineage development, possibly due to its ability to interact with and differentially modulate the activities of other transcription factors that are expressed in lineage- and/or temporally defined patterns.

Zac1 and the Regulation of Cell Cycle Exit: Tumor Suppressor or Oncogene?

Until recently, genes involved in cancer had either been classified as oncogenes or tumor suppressors, depending primarily on their ability to induce or inhibit cell cycle progression, respectively. However, such a binary classification system is incomplete, as several genes display the hallmarks of both tumor suppressor and oncogene depending on cellular context (Baker and McKinnon, 2004). Previously, Zac1 had been implicated as a tumor suppressor based on its ability to induce cell cycle arrest and apoptosis when misexpressed in epithelial cell lines (Spengler et al., 1997; Bilanges et al., 2001), and its inactivation in breast (Bilanges et al., 2001), pituitary (Pagotto et al., 2000), head and neck (Koy et al., 2004), and ovarian carcinomas (Abdollahi et al., 1997a, 2003). However, the widespread expression of Zac1 in dividing precursors throughout the developing neural tube, including the retina, is not consistent with this gene's role in inducing cell cycle exit, as most cells that express Zac1 continue to divide for some time (Alam et al., 2005). More consistent with the expression profile of Zac1, we found that mZac1c could instead promote cell division at an intermediate-stage of retinogenesis (i.e., stage 32), and not at an earlier stage (i.e., stage 28), or a later stage (i.e., stage 40). The stage-specific role of mZac1c in regulating cell cycle progression is analogous to the previously observed stage-specific requirements for the cyclin-dependent kinase inhibitors p27Kip1 and p57Kip2 in regulating cell cycle exit in the retina (Dyer and Cepko, 2000, 2001; Levine et al., 2000), providing additional support for multiple phases of regulation of the cell cycle in the retina.

How can the two apparently contradictory effects of Zac1 on cell cycle progression in Xenopus retina and in cell lines in vitro be reconciled? The answer may come from our knowledge of Zac1′s role as a transcription factor. Specifically, Zac1 acts as a direct transcriptional activator or repressor in some tissues, and acts as a coactivator or corepressor of nuclear receptors and other transcription factors in other cell types, indicating that Zac1 function is highly context-dependent (Huang and Stallcup, 2000; Huang et al., 2001; Hoffmann et al., 2003). Indeed, Zac1 function is reminiscent of the Runx genes, which can also act as transcriptional activators or suppressors depending on cellular context, and can induce ectopic proliferation when either ablated or overexpressed (Cameron and Neil, 2004). Furthermore, two different studies of PTEN, a phosphatase that acts as a tumor suppressor, revealed that tissue-specific ablations of this gene could either reduce or augment the rate of proliferation of neural precursors in the hindbrain, suggesting that the timing of gene deletion is critical (Groszer et al., 2001; Marino et al., 2002). In the future, a better understanding of how Zac1 differentially affects cell proliferation may come from an identification of downstream targets.

Zac1 Function in RGC and Müller Glia Cell Lineages

The timing of cell cycle exit is an important determinant of cell fate choice in the retina (Levine and Green, 2004), and transcription factors and bona fide cell cycle regulators control both processes (e.g., Prox1, Rb, p27Xic1; Ohnuma et al., 1999; Dyer et al., 2003; Chen et al., 2004; Zhang et al., 2004). Accordingly, we found that Zac1 was not only sufficient to influence cell proliferation, but also to specify retinal cell fates. Zac1 misexpression resulted in an expansion of RGC layer and Müller glia cell populations at the expense of most other cell types. While we found that Zac1 misexpression enhanced apoptosis, such a phenomenon does not easily account for the formation of cell clusters, nor does the minimal increase in apoptosis, observed only at an intermediate stage of retinogenesis, likely explain the significant effects of Zac1 on cell fate specification. Of interest, the ability of Zac1 to influence cell fates was apparently stage-specific, since Zac1 initially promoted an expansion of the RGC layer population without affecting Mueller glia cell numbers, and only increased Müller glia cell numbers at a later stage when precursor cells were competent for gliogenesis.

A possible interpretation of our data is, therefore, that Zac1 indirectly promotes the generation of supernumerary Müller glia, which are the last cells to be born, by maintaining retinal progenitor cells in a proliferative mode (i.e., increased proliferation as observed at stage 32) until the environmental signals for Müller glial fate specification are present. A second interpretation is that Zac1 directly influences cell fates but that its activity is context-dependent, raising the question of with which transcription factors Zac1 could interact. A potential clue comes from the observation that Zac1 gain-of-function phenotypes are similar to the Math3;NeuroD double mutant phenotypes, as these mice have increased numbers of RGCs and Müller glia (Inoue et al., 2002). The caveat to this statement is that the supernumerary cells induced by Zac1 in the GCL are not completely specified/differentiated RGCs, suggesting that Zac1 interferes with RGC development. However, it remains possible that Zac1 may negatively regulate Math3 and NeuroD function, possibly by acting as a direct corepressor, akin to its known mode of regulation of a class of nuclear receptors (Huang and Stallcup, 2000), a hypothesis that deserves further study. Of interest, Zac1 misexpression profoundly affected the differentiation of cells in the RGC layer, resulting in morphological and molecular abnormalities, suggesting that Zac1 must be either turned off or down-regulated during early stages of RGC differentiation for this process to occur normally.

Our analyses of Zac1 function involved misexpression of a mouse gene in Xenopus. That transcriptional regulators have conserved functions across vertebrate phyla justified this often used approach (Zuber et al., 2003). We recognize, however, that gain-of-function analyses of mouse genes in Xenopus may not reliably yield the same phenotypes as when the same genes are misexpressed in mouse (Kanekar et al., 1997; Brown et al., 1998; Morrow et al., 1999; Perron et al., 1999; Hatakeyama et al., 2001; Inoue et al., 2002; Moore et al., 2002; Hatakeyama and Kageyama, 2004). At the same time, in all reported cases, mouse genes that bias retinal cell fates or influence other processes such as cell cycle progression in Xenopus retina have some capacity to influence retinogenesis in mouse, even if the cell types affected differ, indicating that the Xenopus system is a reliable indicator that a gene is a key regulator of retinal development. Moreover, with respect to Zac1, in the Xenopus system, deletion of the first two zinc finger domains, which interferes with transactivation properties, also severely attenuated the effects of Zac1Δ2 on cell fate specification. Thus our data strongly indicate that Zac1 functions in a dynamic manner during retinogenesis, likely regulating processes that include proliferation, apoptosis, cell fate specification, and differentiation. In the future, studies of this apparently important multifunctional regulator of retinogenesis will lead to new insights into how the retina is formed.


Bioinformatics, RNA Isolation, RT-PCR, and Generation of Expression Constructs

Eyes dissected from E15.5 mouse embryos were frozen on dry ice, and RNA was extracted by a modified Chomczynski method (Chomczynski and Sacchi, 1987) as described (Mattar et al., 2004). mZac1 was amplified by PCR from retinal cDNA or from an E13.5 embryonic cDNA library (ResGen) using the following primers: 5′ primer: AATCTAGACATGGCTCCATTCCGCTGTCA; 3′ primer: AATCTAGAATCTCCCACAAGAGAAGCACC. For Zac1Δ2, the same 3′ primer was used, and the 5′ primer: AATCTAGACATGGCCACACACTCGCCACA (35 cycles; 94°C/1′, 55°C/1′, 72°C/1′). Amplified cDNA was subcloned into pCS2-MT (gift of D. Turner) for overexpression. A survey of Zac1 ESTs was performed using the UCSC genome browser, and nucleotide comparisons were performed using the NCBI Blast program.

Xenopus Injections

Embryos were obtained from adult Xenopus by using in vitro fertilization. Female Xenopus laevis were stimulated by injection of 800 international units of chorionic gonadotropin (Chorulon; Intervet, Holland), and eggs were collected and fertilized. For blastomere injections, 16-cell stage embryos were dejellied in 2% cysteine (pH 8.0) and injected with plasmid DNA (80 ng/μl), using a borosilicate glass needle pulled on an electrode puller (Sutter Instruments). Pressure injections were performed by using a Picospritzer II (General Valve Company), and targeted the two dorsal blastomeres fate-mapped to give rise to the retina (Moody, 1987). Embryos were allowed to develop in 0.1× Modified Barth's solution (MBS; 8.8 mM NaCl, 0.1 KCl, 0.7 mM MgSO4, 5 mM HEPES [pH 7.8], 25 mM NaHCO3) at 14°C overnight and were subsequently raised in 0.1× Marc's Modified Ringers (MMR; 0.1 M NaCl, 2 mM KCl, 1 mM MgCl2, 2 mM CaCl2, 5 mM HEPES, pH 7.5) at 14–25°C. Embryos were staged as described by Nieuwkoop and Faber (1994).


Xenopus embryos were fixed in 4% paraformaldehyde (PFA)/1×-phosphate buffered saline (PBS) overnight at 4°C. Embryos were then cryopreserved in 30% sucrose/1× PBS overnight and embedded in OCT (Tissue-Tek). Twelve-micrometer cryosections were cut, and sections were washed 2 × 5 min in 1× PBS before blocking, and immunostained using the following primary antibodies: BrdU (5-bromo-2′-deoxyuridine; pretreatment with 2 N HCl, 30 min at 37°C; 1/200; Roche), Islet-1 (1:100; 394D5, Developmental Studies Hybridoma Bank [DSHB]), Kv4.2 (1/50; Alomone), myc (1/500; 9E10, DSHB), NCAM (1/50; DSHB), NF (1:1,000; Chemicon), XEN-1 (1:100; DSHB), GABA (1/3,000; Sigma), Zn-12 (1/100; DSHB), and activated-caspase 3 (1/100; Promega). Primary antibodies were incubated on slides overnight at 4°C and detected using Cy3 (1/500; Jackson ImmunoResearch Laboratories, Inc.) or Alexa488 (1/500; Molecular Probes) -conjugated secondary antibodies.

BrdU Labeling

To label S-phase progenitors in Xenopus, 3 pulses of 5 mg/ml BrdU were injected into the bellies of Xenopus embryos 2 hr before fixation in 4% PFA/1× PBS at 4°C overnight. Xenopus embryos were collected at stages 28, 32, 37/38, and 40.

Microscopy and Cell Counting

Data were collected from three independent experiments. Statistical variation was determined using the standard error of the mean (SEM), and statistical significance was calculated by using either a Student's t-test or an unpaired ANOVA followed by a Student–Newman–Keuls post hoc test.


We thank Dr. William Harris for helpful advice on the manuscript. S.M. and C.S. are Alberta Heritage Foundation for Medical Research (AHFMR) Senior Scholars, S.M. is a Canada Research Chair and C.S. is a Canadian Institutes of Health Research (CIHR) New Investigator. S.M. was funded by a CIHR operating grant, and C.S. was funded by a March of Dimes operating grant. L.M. was supported by a CIHR training grant and J.C.H. by AHFMR and Natural Science and Engineering Research Council studentships.