Zac1 functions through TGFβII to negatively regulate cell number in the developing retina

Background Organs are programmed to acquire a particular size during development, but the regulatory mechanisms that dictate when dividing progenitor cells should permanently exit the cell cycle and stop producing additional daughter cells are poorly understood. In differentiated tissues, tumor suppressor genes maintain a constant cell number and intact tissue architecture by controlling proliferation, apoptosis and cell dispersal. Here we report a similar role for two tumor suppressor genes, the Zac1 zinc finger transcription factor and that encoding the cytokine TGFβII, in the developing retina. Results Using loss and gain-of-function approaches, we show that Zac1 is an essential negative regulator of retinal size. Zac1 mutants develop hypercellular retinae due to increased progenitor cell proliferation and reduced apoptosis at late developmental stages. Consequently, supernumerary rod photoreceptors and amacrine cells are generated, the latter of which form an ectopic cellular layer, while other retinal cells are present in their normal number and location. Strikingly, Zac1 functions as a direct negative regulator of a rod fate, while acting cell non-autonomously to modulate amacrine cell number. We implicate TGFβII, another tumor suppressor and cytokine, as a Zac1-dependent amacrine cell negative feedback signal. TGFβII and phospho-Smad2/3, its downstream effector, are expressed at reduced levels in Zac1 mutant retinae, and exogenous TGFβII relieves the mutant amacrine cell phenotype. Moreover, treatment of wild-type retinae with a soluble TGFβ inhibitor and TGFβ receptor II (TGFβRII) conditional mutants generate excess amacrine cells, phenocopying the Zac1 mutant phenotype. Conclusion We show here that Zac1 has an essential role in cell number control during retinal development, akin to its role in tumor surveillance in mature tissues. Furthermore, we demonstrate that Zac1 employs a novel cell non-autonomous strategy to regulate amacrine cell number, acting in cooperation with a second tumor suppressor gene, TGFβII, through a negative feedback pathway. This raises the intriguing possibility that tumorigenicity may also be associated with the loss of feedback inhibition in mature tissues.


Page 2 of 21
(page number not for citation purposes)

Background
Tissues and organs are genetically programmed to achieve their optimal, mature size, defined by total cell number and individual cellular dimensions. Several regulatory strategies are employed to control cell number, including: direct negative regulators, which inhibit alternative cell fates but permit (or instruct) a primary fate; negative feedback pathways, acting as cell sensors that halt the continued genesis of specific cell types once a feedback signal reaches threshold levels; and cell counting mechanisms, whereby the number of times a progenitor divides before differentiating is genetically determined [1,2]. In the vertebrate retina, negative feedback pathways are used recurrently for cell number control. The retina is composed of one glial and six neuronal cell types that are present in stereotyped proportions in each vertebrate species [3][4][5]. Based on lineage tracing, all retinal cell types are derived from multipotent progenitor cells [6][7][8][9][10][11], although distinct cell lineages likely also exist [1,12]. In mouse, retinal ganglion cells (RGCs), horizontal cells, cone photoreceptors and amacrine cells are primarily generated during the second half of the embryonic period, while rod photoreceptor, bipolar and Müller glial cell production ends on postnatal days (P) 5-6 in the central retina [3]. Differentiated RGCs, amacrine cells and cones secrete signals negatively regulating production of additional cells of that type [13][14][15][16]. However, only signals limiting production of RGCs have been identified, including Sonic hedgehog (Shh) and growth and differentiation factor-11 (GDF11) [17]. GDF11, a transforming growth factor (TGF)β family member, has similar autoregulatory functions in other tissues, including the olfactory epithelium [18] and pancreas [19], while a related molecule, GDF8 (myostatin), negatively regulates skeletal muscle mass [20], suggesting a common role for these cytokines in cell number control.
We identified Zac1 (zinc finger protein that regulates apoptosis and cell cycle arrest or pleiomorphic adenoma gene-like 1 (Plag-l1)) [21] in a screen designed to isolate genes involved in neural fate specification [22]. Zac1 encodes a seven-C 2 H 2 zinc finger protein that acts as a transcriptional activator or repressor [21]. Zac1 is a known tumor suppressor gene and is frequently lost in multiple carcinomas [21]. Zac1 is also maternally repressed through genomic imprinting, a mode of epigenetic control common to many genes regulating embryonic growth. Recently, a Zac1 null mutation was shown to lead to intrauterine growth restriction, consistent with the kinship theory that paternally expressed genes are growth promoting [23]. However, growth retardation was not expected if Zac1 has tumor suppressor properties, promoting cell cycle exit and apoptosis [21,24]. We therefore examined Zac1 function at the cellular level, focusing on the developing retina, where it is robustly expressed [25]. Notably, in our initial cross-species studies in Xenopus, murine Zac1 unexpectedly promoted proliferation [26]. Herein we describe intra-species loss-and gain-of-function assays in mouse that in contrast reveal tumor suppressor-like properties for Zac1 in the retina. Zac1 is required to induce cell cycle exit and apoptosis at late developmental stages, with Zac1 mutant retinae becoming hypercellular, containing supernumary rod photoreceptors and amacrine cells. Strikingly, Zac1 negatively regulates rod and amacrine cell numbers through distinct autonomous and cell nonautonomous (TGFβII-mediated) inhibitory mechanisms, respectively.

Biphasic expression of Zac1 in retinal progenitors and postmitotic cells
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).

Zac1 mutants develop hypercellular retinae containing an ectopic cellular layer
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 inherit- ing 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.

Negative feedback signals are deficient in Zac1 +m/amacrine cells
Our data suggested that the 'stop' or negative feedback signals that normally limit amacrine cell production later in development [13,14] 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 20fold 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 wildtype 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.
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.
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; TGF-βRII-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.

Discussion
The development of a functional retina requires that appropriate numbers of each cell type be generated. Hence, the molecular events that guide cell fate specification and differentiation must be tightly coordinated with those that govern cell number control. Here we demonstrate that the Zac1 tumor suppressor is an essential negative regulator of retinal size, controlling the absolute number of rod photoreceptors and amacrine cells generated during development. Strikingly, Zac1 regulates rod and amacrine cell genesis through distinct cell autonomous and non autonomous mechanisms, respectively ( Figure 9). While Zac1 is a direct negative regulator of a rod photoreceptor fate, it regulates amacrine cell genesis Zac1 regulates TGFβII signaling in the retina Figure 7 Zac1 regulates TGFβII signaling in the retina. (a-f) Co-expression of TGFβRI (green (a-c)) and TGFβRII (green (d-f)) with Ccnd1 (red, proliferating progenitors (b,e)) and Pax6 (red, amacrine cells (c,f)) in E18.5 > 4DIV wild-type retinal explants. (g-l) TGFβII expression in E18.5→4DIV wild-type (green (g-i)) and Zac1 +m/-(green (j-l)) retinal explants co-labeled with Pax6 (red, amacrine cells (i,l)). Arrowheads mark double + cells. Asterisk in (j) marks reduction in onbl/INL expression. (m,n) Expression of pSmad2/3 in E18.5→4DIV wild-type (m) and Zac1 +m/-(n) retinal explants. (o) Western blot analysis of TGFβII, pSmad2/3, total Smad2/3, and β-actin. Asterisks in (o) indicate mutants with reduced expression of TGFβII or pSmad2/3. (p,q) Quantitation of expression levels normalized to β-actin via densitometry for TGFβII (p) and pSmad2/3 (q). TGFβII negatively regulates amacrine cell genesis by controlling the expression of TGFβII, which serves as an amacrine cell negative feedback signal. Zac1 and TGF-βII thus join a growing list of tumor suppressor genes with established roles in retinogenesis (for example, Rb, p53, p27 Kip1 [33,35,40,44,[47][48][49][50][51][52]), but are the first tumor surveillance molecules shown to control neuronal number through a negative feedback or 'cell sensing' mechanism.

Zac1 promotes cell cycle exit and apoptosis in the developing retina
The widespread expression of Zac1 in dividing progenitor cells in the retina (this study) and throughout the developing neural tube [25,[53][54][55] 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 [35,52] and in contrast to Rb mutants, where committed precursors instead fail to exit the cell cycle [33,47,48]. 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 [35,52,57], 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].

Zac1 functions as a direct negative regulator of rod cell fate
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 nonautonomous 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 [59,60]. 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 [21,24]. 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.

Zac1 regulates amacrine cell production cell nonautonomously
Previous studies based on ablation of mature amacrine cells [14] and aggregation of early progenitors with postmitotic 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 [1,13]. 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.

Conclusion
Here we demonstrate that Zac1 is an essential negative regulator of retinal size, controlling the absolute number of rod and amacrine cells generated during development. Strikingly, while Zac1 acts as a direct negative regulator of a rod fate, it negatively regulates amacrine cell genesis via TGFβII-mediated negative feedback inhibition. Zac1 and TGFβII are thus the first tumor surveillance molecules shown to control neuronal number through a negative feedback, 'cell sensing' mechanism. In summary, Zac1 regulates cell number and migration in the developing retina, highly reminiscent of its function in the prevention of tumor formation, suggesting that similar cellular and molecular mechanisms may underlie these processes.

BrdU labeling
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]).

Retinal electroporation
For misexpression, full-length Zac1 cDNA [26] was cloned into a pCIG2 expression vector containing a CMVenhancer/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.

Cell counts and statistical analysis
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 indi-vidual 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.

Additional data files
The following additional data are available with the online version of this paper.