INTRODUCTION

During the development of the eye, in vertebrates and invertebrates, neural progenitors derive from multipotent and proliferative cells (reviewed by Chow and Lang, 2001; Dominguez and Casares, 2005). In the Drosophila eye primordium, the TALE-class homeodomain transcription factor Homothorax (Hth) is expressed in this multipotent population, where it is required to maintain these cells in a proliferative state and to prevent their premature differentiation (Pai et al., 1998; Pichaud and Casares, 2000; Bessa et al., 2002). The homologs of hth in vertebrates are the Meis and Prep (also known as Pknox) gene families (reviewed by Burglin, 1997; Moens and Selleri, 2006). Whereas the expression of Prep genes is widespread in mice and zebrafish, Meis genes show specific transcription patterns in vertebrates, including expression in the developing eye (Ferretti et al., 1999; Toresson et al., 2000; Waskiewicz et al., 2001; Maeda et al., 2002; Zhang et al., 2002; Hisa et al., 2004).

Recent work points to a role for Meis genes in eye development: Meis1 and Meis2 are upstream regulators of Pax6 in the developing lens in chicken and mouse (Zhang et al., 2002), and mouse embryos homozygous for a homeodomain-less Meis1 gene show eye malformations (Hisa et al., 2004). Still, the precise role(s) played by Meis genes during eye development remain(s) unknown. If the parallels in early eye development between flies and vertebrates hold true for Hth/Meis, Meis genes might be involved in stimulating proliferation, or preventing premature differentiation in the optic primordium, or both. Here, we investigated these hypotheses in the zebrafish (Danio rerio).

MATERIALS AND METHODS

Probe preparation, in situ hybridization and immunolabeling

Antisense RNA probes were prepared from cDNAs and labeled with digoxigenin. Specimens were fixed, hybridized and stained as described (Tena et al., 2007).

Fluorescent probes and antibodies

Propidium iodide (PI) was used as nuclear stain; FITC-phalloidin to mark filamentous actin; anti-Islet1 mouse monoclonal antibody labels GCL [36 hours post-fertilization (hpf)] and ganglion cell layer (GCL) plus inner nuclear layer (INL) (48-72hpf) (from DSHB, University of Iowa); rabbit anti-GFP (A11122, Molecular Probes), mouse anti-Myc (MMS 150P, Covance), mouse anti-cleaved Caspase 3 (Cell Signaling Technology). Fluorescent secondary antibodies were from Molecular Probes. Dissected eyes from stained embryos were imaged using a Leica-SP2 confocal system, and data processed with Adobe Photoshop.

In vitro RNA synthesis and microinjection of mRNA and morpholinos

cDNAs were linearized and transcribed as described (Tena et al., 2007). One- to two-cell-stage zebrafish embryos were injected in the yolk with mRNA and/or morpholino (MO) diluted in ∼5 nl of injection solution (10% Phenol Red in DEPC-treated water).

MOs targeting the ATG region of meis1, meis2.2, meis3 and meis4 mRNAs (see Fig. S1A in the supplementary material) were synthesized by GeneTools. We verified the target specificity of meis1- and meis2.2-MOs in Xenopus laevis assays (see Fig. S1B in the supplementary material), and the biological specificity of the meis1-MO by testing its ability to reduce the rhombomere-3 expression of krox20 (also known as egr2 - ZFIN) (see Figs S2 and S5 in the supplementary material).

As controls, we injected similar amounts (8-16 ng) of a control MO directed against the Xenopus tropicalis olig2 gene that shows no match in the zebrafish genome (see Fig. S1 in the supplementary material). The meis3-MO, which has nine and seven mismatches with meis1 and meis2.2, respectively, also served as control MO in some experiments.

Eye phenotype measurements

The polygonal-lasso tool from Adobe Photoshop was used to measure in digital photographs taken with the same magnification, the eye surface area (in pixels) of control and morphant embryos. The volume of each eye was estimated considering it as a hemisphere of radius equal to the radius of a circle with that same area. Measurements from 20 eyes for each condition were compared using a χ² test.

Plasmid constructs

I.M.A.G.E. cDNA clones, from the Lawrence Livermore National Laboratory Consortium, used were: ccnd1 (IMAGE IRALp962K2356Q), c-myc (IRBOp991F125D), meis1 (IRAKp961C08136Q), meis2.1 (IRBOp991C0733D), meis2.2 (IRBOp991D0437D), meis3 (IRALp962E1456Q) and meis4 (MPMGp609N1326Q). pCS2-ccnd1 was generated by inserting the full-length cDNA into EcoRI and XbaI sites of pCS2+. To generate GFP-meis1, MT-meis1, meis1-MT, MT-meis2.2, meis2.2-MT, MT-meis3, meis3-MT, MT-meis4 and meis4-MT constructs, we PCR amplified the corresponding Meis coding regions with the following primer pairs (5′-3′; EcoRI and XhoI sites underlined): GAATTCGATGGCGCAGAGGT and CTCGAGCATGTAGTGCCACTGTCCC for meis1; GAATTCGATGGCGCAAAGGTACGA and CTCGAGCATGTAGTGCCACTGGCC for meis2.2; GAATTCCATGGATA AGAGGTATGAGGAGTT and CTCGAGGTGGGCATGTATGTCAA for meis3; and GAATTCCATGGCGCAACGGTACGA and CTCGAGCATGTAGTGCCACTGACTCTC for meis4.

The PCR fragments were subcloned into pGEMT-Easy (Promega) and sequenced. Meis cDNAs were cloned into pCS2 MT, pCS2p+MTC2 or pCS2eGFP (kindly provided by D. Turner, University of Michigan, USA) to generate N-terminal (Myc-meis) and C-terminal (meis-Myc) Myc-tagged meis or N-terminal GFP-tagged meis1 (GFP-meis1), respectively. To generate the Tol2-GFP-meis1 and Tol2-GFP constructs, we inserted the GFP-meis1 and GFP fragments, respectively, into SalI and SspI sites of Tol2 (pT2KXIG).

Acridine Orange staining

Acridine Orange staining was performed as described (Perkins et al., 2005).

DNA content analysis and flow cytometry

Eyes dissected from 19hpf zebrafish embryos were disaggregated, and PI staining carried out as described (Langenau et al., 2003). DNA content was analyzed on a BD FACSAria and results processed with FloJo software (Tree Star). A χ² test was used for statistical data analysis.

Induction of ectopic expression mosaics

The Tol2 transposon/transposase method of transgenesis (Kawakami et al., 2004) was used with minor modifications. Four- to 16-cell-stage zebrafish embryos were injected in the yolk with 5-12.5 pg of either Tol2-GFP-meis1 or Tol2-GFP constructs, plus 125 pg of transposase-encoding mRNA in a final volume of 5 nl of injection solution. Embryos were cultured at 28.5°C, staged and fixed. Anti-GFP antibody was used to detect the GFP- or GFP-meis1-expressing clones. A stack of confocal z-sections was obtained for each eye analyzed. Three-dimensional reconstruction of the stacks was used to determine the location of the clones.

RESULTS AND DISCUSSSION

meis1 expression is restricted to the undifferentiated and proliferating cells of the early zebrafish eye

Of all five zebrafish Meis genes (meis1, 2.1, 2.2, 3 and 4.1), only meis1 and meis2.2 are expressed during early stages of eye development (Kudoh et al., 2001; Waskiewicz et al., 2001; Zerucha and Prince, 2001; Thisse and Thisse, 2005) (this work). meis1, as monitored by in situ hybridization, or by a YFP insertional reporter inserted close to meis1, was seen to be uniformly transcribed in the eye primordium from 15 to ∼24hpf (Fig. 1A and see Fig. S3 in the supplementary material), a period in which all cells proliferate (Li et al., 2000). After this time, meis1 expression progressively retracted in the retina (Fig. 1B-D,K,L) as the neurogenic wave, marked by ath5 (also known as atoh7 - ZFIN) expression, expands from antero-nasal to posterior-temporal positions (Fig. 1F-H) (Hu and Easter, 1999; Li et al., 2000; Masai et al., 2000). meis1 remained transiently expressed in the ciliary margin zone (CMZ), where the retinal stem population resides (Fig. 1D,M). meis2.2 was also found to be expressed uniformly in early eye primordia, but its expression faded away by 20hpf (see Fig. S3 in the supplementary material). Similar to the situation found in chicken and mouse (Zhang et al., 2002), meis1 was expressed in the prospective lens ectoderm, but was turned off as the lens placode started to thicken (Fig. 1I,J). Therefore, meis1 expression is associated with the undifferentiated, proliferative cells during the early development of the zebrafish eye. In addition, a new wave of Meis gene expression starts in postmitotic neurons at around 36-42hpf (Fig. 1M and see Fig. S3 in the supplementary material). Interestingly, at 4 days post-fertilization (dpf), meis2.2 expression had replaced meis1 at the CMZ.

meis1 is required to promote the G1-to-S transition of the eye primordium cells, and regulates the transcription of cyclin D1 and c-myc

The early expression pattern of meis1 suggests that it plays a role in the proliferative/multipotent cells of the developing eye. To determine what role that is, we knocked down meis1 function using meis1-specific morpholinos (meis1-MO). By the end of embryogenesis, meis1 morphants were severely microphthalmic (>60% of embryos injected with 8 ng of meis1-MO, n=163) (Fig. 2A,B,E,F), with eyes containing fewer cells than controls (Fig. 2C,D). Despite this, meis1-morphant eyes showed apparently normal retinal lamination (Fig. 2C′,D′). The lens was normal or slightly reduced (Fig. 2D and data not shown). The co-expression of meis2.2 and meis1 during optic vesicle stages suggested a possible functional redundancy between these two genes. Nevertheless, injection of meis2.2-MO (8 ng/embryo) caused only mild eye reductions in 22% of the treated embryos (n=299). Furthermore, co-injection of equivalent amounts of meis1- and meis2.2-MOs (4 ng of each MO/embryo) did not significantly enhance the penetrance or severity of the microphthalmia (29%, n=196). These results suggest that meis2.2 does not have a major role during early stages of eye development in the zebrafish. The phenotype observed in meis1 morphants does not appear to be due to an abnormal eye primordium specification. Although we found a slight decrease in pax6b expression in meis1-morphant eyes, as estimated by RT-PCR (see Fig. S4C,D in the supplementary material), the early expression of the eye selector genes pax6b and rx2 (Stigloher et al., 2006) seemed unaffected, when observed by in situ hybridization (see Fig. S4A,B in the supplementary material and data not shown). In agreement with this, we found that the expression of the Pax6 gene eyeless is independent of hth during the development of the Drosophila eye (see Fig. S4E,F in the supplementary material).

To further dissect the mechanisms underlying the observed microphthalmia, we assessed whether meis1 controls the cell cycle. meis1-morphant eyes, at 19hpf, had a significantly higher percentage of cells in G1 phase than control embryos (Fig. 3A,E), indicating a requirement of meis1 in promoting the G1-S transition. Viability of these cells was not compromised, as meis1-morphant eyes did not show a significant increase in the levels of active Caspase 3, or in the vital incorporation of Acridine Orange (not shown).

cyclin D1 (ccnd1) and c-myc (also known as myca - ZFIN) are two major G1 regulators of the cell cycle in vertebrates (Levine and Green, 2004). During the development of the zebrafish eye, cyclin D1 and c-myc are first widely expressed in the optic vesicle, followed by a progressive restriction to the proliferating cells of the neural retina (Thisse and Thisse, 2005; Yamaguchi et al., 2005), a pattern that is reminiscent of that of meis1. In addition, recent work shows that cyclin D1 is required for proliferation in the zebrafish developing retina, as cyclin D1 morphants are microphthalmic (Duffy et al., 2005). The similarity between the patterns of expression of meis1, cyclin D1 and c-myc, and the similar eye phenotypes of cyclin D1 and meis1 morphants, prompted us to ask whether cyclin D1 and c-myc were under meis1 control. Indeed, meis1 morphants showed a dramatic reduction of cyclin D1 and c-myc expression in the eye when compared with control-injected embryos (Fig. 3F-I and see Fig. S5 in the supplementary material). In addition, the co-injection of either cyclin D1 or c-myc mRNAs partially rescued the cell cycle defects of meis1 morphants to levels similar to those obtained by co-injection of GFP-meis1 mRNA (Fig. 3B-E). These results place cyclin D1 and c-myc functionally downstream of meis1 in the control of cell cycle progression in the developing eye. Whether meis1 regulates the transcription of cyclin D1 and c-myc directly or indirectly is unknown.

Maintenance of meis1 expression is incompatible with cell differentiation

In Drosophila, hth not only promotes proliferation in the eye primordium, but forced maintenance of its expression results in a delay or block of retinal differentiation (Pai et al., 1998; Pichaud and Casares, 2000; Bessa et al., 2002). Similarly, in the early zebrafish eye, meis1 expression is found in undifferentiated cells but is turned off as neurogenesis advances (Fig. 1). To test whether maintaining meis1 expression is incompatible with retinal differentiation, we analyzed the distribution of clones of cells expressing either GFP or GFP-tagged-Meis1 in developing retinas, prior to and after the initiation of neuronal differentiation (Fig. 4 and see Fig. S6 in the supplementary material). Differentiation was followed using the GCL marker islet1. When analyzed between 24 and 30hpf, a stage at which most of the retina is undifferentiated, all GFP- and 80% of GFP-Meis1-expressing clones spanned the whole width of the neuroepithelium (n=57 and 46, respectively; Fig. 4A,D). Later in development, when retinal differentiation is ongoing and layering becomes apparent, most GFP clones appeared in the central retina and contained both Islet1-expressing and non-expressing cells (90%) (Fig. 4B,C), whereas only a few (7%) were found in the CMZ (n=41). By contrast, of the GFP-Meis1 clones located in the central retina (72%, n=39), none contained Islet1-positive cells at this stage (Fig. 4E). In addition, a large portion of these Meis1-expressing clones (28%, n=39) was found in the CMZ (Fig. 4F). The fact that Meis1-expressing cells were always found in undifferentiated regions of the neuroepithelium, leads us to conclude that maintenance of meis1 expression in the first 48 hours of eye development is incompatible with neuronal differentiation.

Our results indicate that, during early eye development, meis1 shares two roles with its fly homolog, hth. First, meis1 is required to maintain proliferation of the multipotent cells of the early eye, by promoting the G1-to-S transition of the cell cycle. Mechanistically, meis1 regulates the transcription of at least two potent cell-cycle activators: cyclin D1 and c-myc. Second, ath5 follows receding meis1 expression in a similar fashion as in Drosophila, where hth expression retracts as the atonal-expressing differentiation wave advances. This finding is in accordance with a model in which the expression of meis1 has to be downregulated to allow further differentiation of the fish retina, and agrees with our results that the sustained expression of meis1 is incompatible with neural differentiation. Similar results have been obtained in chicken and mouse by Heine and co-workers (Heine et al., 2008). Although retinal lamination in meis1 morphants is not grossly affected, we cannot rule out specific effects on the specification and/or differentiation of specific retinal cell types, as meis1, together with meis2.1 and meis2.2, is redeployed in postmitotic cells of the ganglion cell and inner nuclear layers (Fig. 1M and see Fig. S3 in the supplementary material).

The expression of meis1 in the CMZ, and the fact that forcing meis1 expression results in the localization of the expressing cells to the CMZ, suggest that meis1 might function in specifying the retinal stem cells of the zebrafish. In this regard, it is interesting to note that meis1 expression resembles that of Pax6, a previously described retinal progenitor transcription factor (Raymond et al., 2006) (reviewed by Amato et al., 2004). In Drosophila, previous results showed that hth and eyeless are co-expressed in the undifferentiated domain and that their products might directly interact in vivo (Bessa et al., 2002). All these results seem to indicate that a common molecular mechanism to maintain a multipotent stem-like state exists during eye development in vertebrates and invertebrates.

In addition to controlling several developmental processes, Meis genes are overexpressed in an increasing number of cancer types (Lawrence et al., 1999; Segal et al., 2004; Geerts et al., 2005; Dekel et al., 2006). Therefore, the identification of functional targets of the Meis genes involved in the maintenance of the undifferentiated and proliferative state during normal development, such as cyclin D1 and c-myc, is likely to be instrumental in deciphering the mechanisms underlying Meis-associated tumors.

Supplementary material

Supplementary material for this article is available at http://dev.biologists.org/cgi/content/full/135/5/799/DC1