Timing the Generation of Distinct Retinal Cells by Homeobox Proteins

The reason why different types of vertebrate nerve cells are generated in a particular sequence is still poorly understood. In the vertebrate retina, homeobox genes play a crucial role in establishing different cell identities. Here we provide evidence of a cellular clock that sequentially activates distinct homeobox genes in embryonic retinal cells, linking the identity of a retinal cell to its time of generation. By in situ expression analysis, we found that the three Xenopus homeobox genes Xotx5b, Xvsx1, and Xotx2 are initially transcribed but not translated in early retinal progenitors. Their translation requires cell cycle progression and is sequentially activated in photoreceptors (Xotx5b) and bipolar cells (Xvsx1 and Xotx2). Furthermore, by in vivo lipofection of “sensors” in which green fluorescent protein translation is under control of the 3′ untranslated region (UTR), we found that the 3′ UTRs of Xotx5b, Xvsx1, and Xotx2 are sufficient to drive a spatiotemporal pattern of translation matching that of the corresponding proteins and consistent with the time of generation of photoreceptors (Xotx5b) and bipolar cells (Xvsx1 and Xotx2). The block of cell cycle progression of single early retinal progenitors impairs their differentiation as photoreceptors and bipolar cells, but is rescued by the lipofection of Xotx5b and Xvsx1 coding sequences, respectively. This is the first evidence to our knowledge that vertebrate homeobox proteins can work as effectors of a cellular clock to establish distinct cell identities.


Introduction
Different types of neurons are generated at predictable times in several developing brain structures [1][2][3]. Although the molecular machinery that links a type of nerve cell to its time of generation has been investigated in the fruit fly [4][5][6], little is known in higher animals. In the vertebrate retina, pluripotent progenitor cells generate the six main types of retinal neurons (ganglion, horizontal, cone, amacrine, rod, and bipolar cells) following an evolutionarily conserved time schedule [1]. This observation suggests that a molecular machinery has been selected to ensure tight coordination between cell birth date (that is the time of exit from the cell cycle) and the specification of a given neuronal cell fate [7].
Changes of retinal cell-fate competence (that is, the capability to generate one type of retinal cell rather than another) are controlled in time and space by the activity of proneural bHLH transcription factors [8]. Moreover, these alone are not sufficient to specify distinct cell fates [9], and several pieces of evidence suggest that they act in concert with homeobox gene products, which seem to refine their action to generate different cell types [10]. A number of homeobox genes were found to be necessary and/or sufficient to establish retinal cell identity: prox1 is both necessary and sufficient for the generation of horizontal cells [11]; Xbh1 promotes ganglion cells [12]; the otx-like crx [13] and otx5 [14,15] support the generation and/or maintenance of photoreceptors; vsx1 [16], chx10/vsx2 [17], and otx2 [14] sustain the production of bipolar neurons. While these data demonstrate the crucial role of homeobox genes in retinal cell identity, they do not address the question of how the neurogenetic timing is controlled. A critical question is therefore when, where, and how retinal homeobox genes are activated during retinal neurogenesis.
We recently observed that long-lasting cell cycle progression (and consequently a late cell birthday) is sufficient to generate late retinal cell types such as rods and bipolar cells [18]. Accordingly, the inhibition of cell cycle progression greatly enhances the capability of the retinal bHLH gene Xath5 to support the generation of ganglion cells, which are the earliest-generated retinal cells, at the expense of bipolar cells, which are the latest-generated neurons [19]. These observations suggest that the activation of homeobox genes that are crucial for late retinal cell types may be linked to cell cycle progression rather than to absolute time. Notably, a similar mechanism occurs in Drosophila, in which the sequential expression of the transcription factors that control different neuronal identities requires cytokinesis [4].
The aim of this work is to provide evidence that the sequential activation of retinal homeobox genes depends on a cellular clock that establishes the cell birth dates of distinct retinal cell types. Here we report that the three Xenopus homeobox genes Xotx5b, Xvsx1, and Xotx2 are translationally regulated with a timing that parallels that of the generation of photoreceptors (Xotxb5) and bipolar cells (Xvsx1 and Xotx2) and that their translation depends on cell cycle progression.
Moreover, we show that the block of cell cycle progression severely affects the generation of photoreceptors and bipolar cells, whereas Xotx5b and Xvsx1 proteins can overcome this effect. Our results confirm the importance of a cellular clock in establishing distinct cell fates and draw attention to the translational control of homeobox genes as a mechanism to regulate the neurogenetic timing in the Xenopus retina.

Results/Discussion
In the Xenopus retina, Xotx2 is both necessary and sufficient for the generation of bipolar cells [14]. Moreover, while also vsx1 [16] and chx10/vsx2 [17] are necessary in other vertebrates, chx10 is not sufficient to support the bipolar fate in Xenopus [20]. We recently isolated the Xenopus homolog of vsx1, Xvsx1, which is expressed by retinal progenitors and, in the mature retina, by bipolar cells (D'Autilia et al., unpublished data). To assay the ability of Xvsx1 to support the generation of bipolar cells, we lipofected the Xvsx1 coding sequence into stage (st.) 17-18 embryonic optic vesicles and compared the proportion of Xvsx1-lipofected cells to controllipofected cells at the stage of mature embryonic retina (st. 42; Figure 1). As reported in Figure 1B and 1C, Xvsx1 lipofection significantly increases the proportion of bipolar cells and decreases that of photoreceptors compared to control lipofection ( Figure 1A and 1D). Thus, in the Xenopus retina, Xotx5b supports photoreceptor differentiation [14], while Xotx2 and Xvsx1 support bipolar cell differentiation.
Photoreceptors (namely rods) and bipolar cells are the latest-generated retinal neurons both in Xenopus [9,21] and in mammals [1]. With the idea that Xotx5b, Xotx2, and Xvsx1 were sequentially activated during retinal neurogenesis, matching the time of photoreceptor and bipolar cell generation, we examined their spatiotemporal pattern of expression. We observed that Xotx5b, Xotx2, and Xvsx1 are strongly regulated at post-transcriptional level, both in time and space. At midretinal neurogenesis (st. 34 [22]), the mRNAs of these three genes show a similar widespread pattern of expression, but only the Xotx5b protein is detectable in few apical nuclei ( Figure 2). Xvsx1 protein detection starts at late-retinal neurogenesis (st. 37), when the mRNAs of the three genes begin to segregate into specific retinal domains ( Figure 2). Xotx2 protein is detectable at high levels from st. 38-39 (not shown) onward. The patterns of protein and mRNA expression are further refined in mature embryonic retinas (st. 42). At this time, Xotx2 mRNA identifies most bipolar cells [14], Xotx5b mRNA marks photoreceptors and a sub-population of bipolar cells [14], and Xvsx1 mRNA labels bipolar cells (D'Autilia et al., unpublished data). At the protein level, Xotx2 and Xvsx1 are detectable in bipolar cells, whereas Xotx5b is visible in photoreceptors but not in bipolar cells. The peculiar post-transcriptional regulation of Xotx5b, Xvsx1, and Xotx2 is maintained in the ciliary marginal zone (CMZ), a proliferating region that, in the mature retina of fishes and amphibians, continues producing new retinal cells and recapitulates all the embryonic developmental steps [23]. Notably, CMZ analysis indicates that Xotx5b, Xvsx1, and Xotx2 proteins start to be detectable in early post-mitotic cells ( Figure S1).
What are the mechanisms controlling protein expression? We found that cis-acting signals in the 39 UTR of Xotx5b, Xvsx1, and Xotx2 mRNAs are sufficient to regulate their pattern of translation. We constructed sensors [24] in which the 39 UTR of each gene was placed downstream of green fluorescent protein (GFP) cDNA (see Materials and Methods). We lipofected these sensors into the optic vesicles and analyzed the distribution of both the GFP mRNA and protein produced by the sensor in single lipofected cells. Figure  The inhibition of GFP translation driven by these UTRs in specific cell types is statistically significant, as reported in Figure 3B. Recently, extensive bioinformatic analyses have shown that at least 20% of the vertebrate genes display in their 39 UTR a family of highly conserved short regions [25], most of which are complementary to a newly discovered class of regulatory short RNAs called microRNAs (miRNAs). Accordingly, cis-acting signals controlling protein translation through the binding of a miRNA have been found in the 39 untranslated region (UTR) of Hox genes [26]. Using an in silico approach, we found that the 39 UTR of Xotx5b, Xvsx1, and Xotx2 contains candidate miRNA domains for 42 distinct miRNAs, four of which are shared by all the three 39 UTRs (Table S1). These miRNA domains are widely dispersed in the 39 UTR of each gene. Whereas such miRNA domains could reveal a functional relevance, specific RNA-binding proteins might also be involved in regulating the translation of retinal Cell types were identified as described [14]. Error bars indicate standard error of the mean. Xvsx1 misexpression increases the proportion of bipolar cells (from 33% of control to 55%, student's t-test, p ¼ 0.000043), mainly at the expense of photoreceptors (from 29% to 14%, student's ttest, p ¼ 0.000011). DOI: 10.1371/journal.pbio.0040272.g001 homeobox genes, since several of them have been shown to control the development and plasticity of the central nervous system, including the retina, in both Drosophila and vertebrates [27][28][29]. Thus, we found that the 39 UTRs of Xotx5b, Xvsx1, and Xotx2 are cis-acting regulators of translational repression, but the molecular nature of the trans-acting repressor(s) remains to be established.
In addition to the role of the 39 UTRs in patterning protein translation, we analyzed their effects on the timing of sensor translation by time-lapse imaging (Figure 4). At st. 30 (assumed as time 0 of imaging), almost all control-lipofected retinas already show GFP-positive clones. The onset of GFP detection in sensor-lipofected retinas is delayed and parallels the onset of detection of the corresponding proteins, peaking at 12 h (st. 35) for Xotx5b sensor, 16 h (st. 37) for Xvsx1 sensor, and at 24 h (st. 39) for Xotx2 sensor. These results show a correlation between the translational onset of a sensor and that of the corresponding gene. An obvious question is whether such timing associates with the cell birth dates of photoreceptors and bipolar cells. By BrdU labeling from st. 30, 34, and 37 and analyses of the BrdU-positive cells in mature retinas, we evaluated the proportion of dividing progenitors fated to generate a given cell type ( Figure S2). The proportion of dividing photoreceptor progenitors drops from 68% (6 2.6% standard error of the mean) at st. 30 to 24% (6 3.3% standard error of the mean) at st. 34, when Xotx5b protein is first detectable. The percentage of proliferating bipolar progenitors falls from 63% (6 5.1% standard error of the mean) at st. 34 to 7% (6 1.2% standard error of the mean) at st. 37, when Xvsx1 protein is first detected, followed soon after by Xotx2 protein detection (st. [38][39]. Thus, there is a temporal correlation between the translational onset of the three genes, the translational onset of sensors, and the cell birth dates of photoreceptor and bipolar cells. We investigated the role of cell cycle progression in the translational control. We found that Xotx5b, Xvsx1, and Xotx2 mRNAs require progressively increasing times of cell cycle progression to be efficiently translated. To establish this point, we blocked cell cycle progression by hydroxyurea/ When detectable, the pattern of expression is comparable to that of control retinas (compare Figure 5 to st. 42 in Figure 2). Although impairing proper retinal lamination, HUA does not interfere with neural cell differentiation [30]. Accordingly, our analysis shows that retinas treated with HUA can express both general neuronal (neurotubulin) and glial (R5 antigen [31]) cell markers ( Figure S3). Treatment from st. 30 does not affect expression of markers for ganglion cells (hermes [32]), horizontal cells (prox1 [11]), amacrine cells (tyrosine hydroxilaseþ, GABAþ, 5-HTþ), and photoreceptors (interphotoreceptor retinoid-binding protein [IRBP] gene [IRBP] [33]).
However, treatment from st. 25 dramatically decreases the expression of IRBP, as well as of amacrine markers (unpublished data).
Is an early block of cell cycle progression sufficient to inhibit the translation of Xotx5b, Xvsx1, and Xotx2 in a cellautonomous way? To test this hypothesis, we inhibited cell cycle progression of single retinal progenitors in a normal environment by the lipofection of the cell cycle inhibitor Xgadd-45c [34,35] (Figure 6). When overexpressed in medaka fish early blastula, Xgadd-45c favors cell cycle exit in G1 [34]. In the Xenopus retina, it is directly induced by Xath5 and expressed by retinal progenitors about to exit from the cell cycle [36]. Misexpression of Xgadd-45c inhibits retinal cell divisions: BrdU injected at st. 33-34 is detected at st. 42 in fewer Xgadd-45c-lipofected cells (19% 6 3.4% standard error of the mean) compared to control-lipofected cells (41% 6 0.79% standard error of the mean; p ¼ 0.001, student's t-test). Significantly, Xgadd-45c-lipofected cells translate Xotx5b, Xvsx1, and Xotx2-co-lipofected sensors less efficiently than cells lipofected with the sensor alone ( Figure 6F). Moreover, the number of Xgadd-45c-lipofected cells that express Xotx5b, Xvsx1, and Xotx2 proteins is considerably lower than that of control cells (not shown). In addition, Xgadd-45c misexpression decreases the proportion of photoreceptor and bipolar lipofected cells compared to control ( Figure 6A-E and 6G).
We propose that it is the decreased production of Xotx5b and Xvsx1 proteins after the block of cell cycle progression that causes the decrease of photoreceptor and bipolar cells. If such is the case, then the effect of Xgadd-45c should be reversed by Xotx5b and Xvsx1 proteins. In fact, the colipofection of the Xotx5b coding region (without 39 UTR) with Xgadd-45c raises the fraction of photoreceptors from 19% (6 1.3% standard error of the mean) to 63% (6 0.9% standard error of the mean; p ¼ 0.0000025, student's t-test) and that of Xvsx1 increases the proportion of bipolar cells from 18% (6 2.2% standard error of the mean) to 58% (6 2.5% standard error of the mean; p ¼ 0.00002, student's t-test) (Figure 7).
On the opposite, the co-lipofection of Xotx5b or Xvsx1 constructs, including the corresponding 39 UTRs, is drastically less effective in rescuing the proportion of photoreceptors cells (Xotx5b coding þ 39 UTR: 31% 6 2.1% standard error of the mean) and bipolar cells (Xvsx1 coding þ 39 UTR: 29% 6 1.8% standard error of the mean) compared to the co-lipofection of the coding sequences alone ( Figure S4). Indeed, the 39 UTR included in these constructs is the same as that which is responsible for the significant decrease of translation when assayed by GFP sensors both in normal (Figures 3 and 4) and in Xgadd-45carrested cells ( Figure 6F). According to the antiproliferative effect elicited by Xgadd-45c, the size of clusters of cells colipofected with Xgadd-45c and Xotx5b ( Figure 7B), or with Xgadd-45c and Xvsx1 ( Figure 7D), was always smaller than the size of clusters of cells lipofected with only Xotx5b (Figure 7A), or with Xvsx1 ( Figure 7C). Remarkably, the majority of bipolar cells generated by the co-lipofection of Xgadd-45c and Xvsx1 have an earlier cell birth date than control-lipofected bipolar cells, as indicated by BrdU-incorporation experiments (not shown). These results indicate that the overexpression of Xotx5b and Xvsx1 proteins can bypass a cell cycle-dependent cellular clock that sets the generation of photoreceptors and bipolar cells. However, these proteins are unlikely elements of the cell clock machinery, because they are not expressed at a detectable level in cycling cells ( Figure S1). Rather, our data suggest that they are downstream effectors of this clock, which could set the time of neuronal generation by translational inhibition.
How would this cellular clock measure the time to set the generation of late cell types? Previous studies in cortical development [37] and in rat retinal development [38] have shown that the cell cycle length of neural progenitor cells increases over time. Because these progenitors generate different cell types over time, this suggests a correlation between cell cycle length and cell fate. By a labeling-index (LI) analysis [39], we estimated that the cell cycle length of midneurogenesis retinal progenitor cells (st. 30) is 5.1 h (6 1.3 h), whereas that of later progenitors (st. 34) is 8.1 h (6 0.6 h; Figure S5A).
To assay a possible functional correlation between cell cycle length and cell fate, we shortened the cell cycle of late progenitor cells by lipofection of the cell cycle regulator E2F, which is necessary for cell cycle progression after midblastula transition [40]. At st. 34, cells lipofected with XE2F have an estimated cell cycle length of 5.5 h (6 1.2 h), whereas cdk2/cyclinA2-lipofected cells show a cell cycle length (7.8 6 0.9 h) that is comparable to that of non-lipofected cells of the same age ( Figure S5B). Nonetheless, more XE2F-lipofected cells (48% 6 2% standard error of the mean) and cdk2/ cyclinA2-lipofected cells (45% 6 0.1% standard error of the mean) are cycling than control-lipofected cells (23% 6 0.1% standard error of the mean) at this stage ( Figure S5B). Thus, both XE2F and cdk2/cyclinA2 lipofection delay the exit from the cell cycle, but only XE2F lipofections shortens cell cycle length. Remarkably, Xotx2 translation at st. 40, soon after its normal onset, is considerably inhibited by XE2F lipofection  [22] (which corresponds to time 0). To better visualize GFP, pigmentation was abolished as described [43]. Each micro-photograph shows the entire area of a lipofected eye and is focused on lipofected cells of the neural retina. Red arrows point to lipofected clones of cells. (B) Statistical analysis of 68 records. Bars express the proportion of lipofected retinas in which GFP was first detectable at a given time. Number of retinas examined is indicated by n. DOI: 10.1371/journal.pbio.0040272.g004 compared to control, whereas it is slightly increased by cdk2/ cyclinA2 lipofection ( Figure 8A-8D).
According to this observation, XE2F-lipofected cells generate significantly fewer bipolar cells than control-lipofected cells, whereas cdk2/cyclinA2 lipofection increases their proportion ( Figure 8E-8G). Since both cdk2/cyclinA2 and XE2F-lipofected cells proliferate longer than control cells, their different cell fates cannot be explained in terms of absolute time spent as cycling cells. Rather, we speculate that a long cell cycle is necessary during the last cell division(s) of a progenitor to translate Xotx2 protein after cell cycle exit and to eventually differentiate as a bipolar cell. These results,  together with the natural lengthening of cell cycle observed during retinogenesis, support the idea that the retinal cell clock would measure the cell cycle length, rather than time, to set the generation of late retinal cells.

Conclusion
We showed that the Xotx5b, Xvsx1, and Xotx2 mRNAs are sequentially translated during retinal neurogenesis. Translational control accounts both for the timing of activation and the cell-specificity of expression of the three genes. Our results show that cell cycle progression is necessary to sequentially remove the translational inhibition of Xotx5b, Xvsx1, and Xotx2. Consequently, homeobox proteins (Xotx5b, Xvsx1, or Xotx2) produced by a retinal post-mitotic cell depend on the cell birth date. Since these three homeobox proteins are crucial for establishing different cell types, this explains in molecular terms why cells that have different birth dates become different retinal neurons.
Our observations indicate that a cellular clock depending on cell cycle progression sets the generation of photoreceptors and bipolar cells and that Xotx5b, Xvsx1, and Xotx2 proteins are downstream effectors of such a clock. Although our observations suggest that this clock measures cell cycle length and that translational inhibitors are part of the clock machinery, its molecular nature is at present unknown. It is now crucial to extend these findings to other retinal cell-fate genes, to dissect the mechanisms of translational inhibition and to find out how cell cycle progression can remove the translational inhibition over developmental time.

Materials and Methods
Embryonic developmental stages were evaluated as described by Nieuwkoop and Faber [22]. To immunodetect Xvsx1, Xotx5b, and Xvsx1 proteins, immunoaffinity-purified polyclonal antibodies were generated in rabbit by PRIMM SRL. Synthetic peptides (three for each of Xvsx1 and Xotx5b, and two for Xotx2), corresponding to 15aalong regions outside the homeobox of the predicted protein sequence, were used as immunogens. Antibody specificity was first assayed by Western blot analysis. Immunostaining of HEK 293T cells, transfected with the coding sequence of the three genes, confirmed the specificity of all three antibodies.
In time-lapse experiments, lipofected retinas were imaged every 4 h, starting from st. 30 (considered as time 0). To avoid pigmentation, embryos were treated with 0.003% N-phenylthiourea (Sigma, St. Louis, Missouri, United States) [43] until imaging. After hatching and shortly before imaging (st. 28), embryos were embedded into soft agar (1%, in 0.1X MMR buffer) in 24-well plates, with their lipofected eyes up. Images of the lipofected eyes were obtained using epifluorescence stereomicroscopy (Nikon SMZ1500 [Tokyo, Japan]) connected with a digital camera (Photometrix CoolSnap, Roper Scientific, Trenton, New Jersey, United States). After imaging, some of the lipofected eyes were fixed, sectioned, and analyzed to confirm the nature of the transfected cells. Figure S1. Xotx5b, Xvsx1, and Xotx2 Expression in the CMZ Comparison between Xotx5b, Xvsx1, and Xotx2 mRNA or protein detection (red) and BrdU-positive cells (green) in the CMZ of st. 42 retinas, after an 8-h BrdU pulse. Since this time of incorporation corresponds to the average cell cycle length of a late embryonic retinal progenitor (see Figure S5), the region of green-labeled cells reasonably excludes the majority of post-mitotic cells. From its most marginal aspect (M) towards the central side of the retina (C), the CMZ recapitulates the different stages of embryonic retinal neurogenesis [23], with more marginal cells earlier (less mature) than more central ones. Green arrows point at the central boundary of BrdU immunodetection, and red arrows indicate the marginal border of mRNA/protein detection. Whereas mRNA detection of the three genes always largely co-localizes with BrdU-labeled cells, none of the cells expressing detectable levels of the corresponding protein contains BrdU. These data indicate that Xotx5b, Xvsx1, and Xotx2 start to be translated at a measurable level in post-mitotic cells. Found at DOI: 10.1371/journal.pbio.0040272.sg001 (743 KB JPG).  Number of counted cells is indicated by n; double asterisk indicates p 0.01; triple asterisk indicates p 0.001 (student's t-test); error bars: standard error of the mean. The lipofection of cdk2/cyclinA2 and XE2F increases and decreases, respectively, the proportion of bipolar cells compared to control. The decrease of photoreceptors after cdk2/cyclinA2 lipofection is due to a reduction of cones [18]. DOI: 10.1371/journal.pbio.0040272.g008 morphology, position in the retinal layers, and expression of markers. The cell birth dates of the different Xenopus retinal neurons are partially overlapping. Nonetheless, rods among photoreceptors [21] and bipolar cells show the latest cell birth dates. A substantial proportion of their progenitors are still dividing (being BrdUpositive) at st. 34 (24% photoreceptors and 63% bipolar). At this stage, only a few ganglion, horizontal, and amacrine progenitors are still dividing (6%, 5%, and 12%, respectively). Bipolar cells are the latest neurons, as the majority of them (63%) are still dividing at st. 34. Found at DOI: 10.1371/journal.pbio.0040272.sg002 (820 KB JPG). Figure S3. Effects of HUA Treatment on Retinal Histogenesis Retinal sections of st. 42 embryos, treated with HUA (150 lM hydroxiurea, 20 lM aphidicoline) from st. 30, compared to control. In situ hybridization of mRNAs (neurotubulin, hermes, prox1, and IRBP) are detected with Fast Red, and antibodies (anti-R5, amacrine antibodies panel) are immunodeteced with Oregon green-conjugated secondary antibody. According to Harris et al. [30], HUA blocks cell proliferation in 4 h from the beginning to the end of treatment, as detected by BrdU-incorporation assay and immunodetection of mitotic cells with the phosphorylated form of Histone3 (not shown). The treatment reduces retinal size but does not impede terminal cell differentiation, as shown by the expression of the Mü ller glial marker R5 [31] and neurotubulin, the staining of which in treated embryos is comparable to control. Immunostaining of the neuronal marker acetylated tubulin (Sigma T6793; 1:1,000) confirmed the observation obtained by in situ hybridization with a neurotubulin probe (not shown). The pattern of neurotubulin and Mü ller glial staining indicates that retinal layering is compromised. This happens, even more severely, when treating from earlier stages (not shown, compare to Harris et al. [30]). The expression of markers for ganglion cells (hermes) and horizontal cells (prox1) is not affected by treatment. The expression of markers for amacrine cells (amacrine antibodies panel as in Figure 1: anti-5-HT, anti-GABA, and anti-tyrosine hydroxilase) and photoreceptors (IRBP) is often reduced but is still detectable with a pattern similar to that of control in all the examined embryos. Treatment from st. 25 strongly reduces IRBP and amacrine markers, but allows the expression of hermes and prox1 (not shown).   In wt retinas, the proportion of cycling cells, expressed as 10 h LI, is higher at st. 30 (32% 6 0.34% standard error of the mean) than at st. 34 (23% 6 0.1% standard error of the mean). This proportion, at st. 34, is even higher in cdk2/cyclinA2-lipofected cells (45% 6 0.1% standard error of the mean) and in XE2F-lipofected cells (48% 6 2% standard error of the mean), which both delay the exit from the cell cycle. In wt retinas, the average cell cycle length, as evaluated by Tc value, increases from st. 30 (Tc ¼ 5.1 6 1.3 h) to st. 34 (Tc ¼ 8.1 6 0.6 h). Notably at st. 34, XE2F lipofection significantly reduces Tc (Tc ¼ 5.5 6 1.2 h) compared to control cells of the same age (Tc ¼ 8.1 6 0.6 h), whereas Tc is not significantly affected by cdk2/cyclinA2 lipofection (Tc ¼ 7.8 6 0.9 h). The changes in Tc observed among different types of cells are poorly affected by Ts, which ranges from 0.9 to 1.4. Found at DOI: 10.1371/journal.pbio.0040272.sg005 (1.7 MB JPG).

Supporting Information
Table S1. In Silico Screening for Candidate miRNA Domains Table S1 shows X. laevis (Xla-mir-), Danio rerio (dre-let/mir-), and Homo sapiens (has-mir-) putative miRNA domains present in the 39 UTR of Xotx5b, Xvsx1, and Xotx2. Among the 30 X. laevis mature miRNAs so far isolated, 27 (90%) show perfect sequence homology with miRNAs of other species [44]. Because of the low number of annotated Xenopus miRNAs and the extreme evolutionary conservation of the mature miRNA sequence, it was reasonable to search also for heterologous domains in the 39 UTR of the three genes. The MIRanda software [45] was used to screen among the miRNA sequences from the three species annotated in the Sanger miRNA registry (http://microrna. sanger.ac.uk). Only results with energy values lower than À20.00 kcal/ mol and score values higher than 100 for at least one of the three UTR are shown. These two thresholds were used in association with gap-open and gap-elongation parameters À8.0 and À2.0, respectively, to ensure high stringency [46]. Sites columns report the number of candidate domains in the corresponding 39 UTR. A total number of 42 different miRNAs show in silico high binding affinities for the 39 UTR of Xotx5b (n ¼ 20), Xvsx1 (n ¼ 28), and Xotx2 (n ¼ 15), four of them (in bold), sharing sites for all three UTRs. Interestingly, two of these shared miRNAs (dre-mir-34 and dre-mir-432) show multiple domains in Xvsx1 and Xotx2 but not in Xotx5b UTR. miRNA domains are reasonably well dispersed over the three UTR sequences in all three genes under consideration. No obvious nucleotide conservation was found among the three 39 UTRs. The selected miRNAs are candidates for the translational repression of Xotx5b, Xvsx1, and Xotx2. However, the presence of the human and D. rerio-selected miRNAs also in Xenopus, as well as the expression of all selected miRNAs in developing retina and their hypothetical role in development, remain to be validated. Found at DOI: 10.1371/journal.pbio.0040272.st001 (20 KB XLS).