Genetic rescue of cell number in a mouse model of microphthalmia:interactions between Chx10 and G1-phase cell cycle regulators

The proliferative expansion of the neuroepithelium is a crucial process in the development of the vertebrate central nervous system (CNS). Not only must sufficient numbers of progenitor cells be generated for neuronal and glial differentiation, but the correct cell number ratios among different CNS tissues must also be attained. Disruptions in the number of CNS progenitor cells result in malformations that seriously impair or eliminate CNS function(Walsh, 1999).

Surprisingly little is known about the molecular mechanisms that regulate region-specific proliferation in the CNS. One reason for this is that cell number variation between different CNS tissues is also dependent on processes that are distinct from proliferation. For example, although regulation of proliferation is an essential aspect of both the development and evolution of the mammalian cerebral cortex (Kornack,2000; Rakic and Caviness,1995), identifying the mechanism of proliferation control in the cortex is complicated by intricate patterning(Monuki and Walsh, 2001;Redies and Puelles, 2001;Sur and Leamey, 2001), cell migration (Gleeson and Walsh,2000, Hatten,1999; Maricich et al.,2001; Ross and Walsh,2001) and cell death (Blaschke et al., 1996; de la Rosa and de Pablo, 2000; Voyvodic,1996).

We have focused on the mechanism of proliferation control in the mammalian neural retina because cell proliferation is a primary determinant of retinal size and cell number. In contrast to other regions of the CNS, cell migration(Stone and Dreher, 1987;Watanabe and Raff, 1988) and cell death (Beazley et al.,1987; Voyvodic et al.,1995; Young, 1984)do not contribute significantly to the size of the total cell population during retinal development. Both processes occur late and involve relatively few cells. Like other areas of the CNS, however, the mammalian neural retina undergoes a massive expansion in cell number by proliferation. From embryonic day 14 (E14) until postnatal day 8 (P8; 16 days), total retinal cell number increases ∼400 fold, from 60,000 cells to 25 million cells in the rat(Alexiades and Cepko, 1996). Furthermore, this expansion in cell number is region specific. The neural retina and retinal pigmented epithelium (RPE) are adjacent tissues, and are both patterned from the optic vesicle. However, the total cell number of the neural retina is much larger than that of the RPE, due primarily to differential regulation of proliferation. Evidence of this kind suggests that tissue-specific regulators of proliferation must exist in different regions of the CNS.

Within the eye, the homeodomain-containing transcription factor Chx10 and its orthologs in fish (Vsx2, Alx1), chicken (Chx10-1), cow and humans are exclusively expressed in the neural retina and adjacent ciliary margin. Within the retina, Chx10 is expressed in retinal progenitor cells(RPCs) throughout the period of proliferation(Barabino et al., 1997;Belecky-Adams et al., 1997;Chen and Cepko, 2000;Levine et al., 1994;Levine et al., 1997a;Liu et al., 1994;Passini et al., 1997). As RPCs become postmitotic and differentiate, Chx10 expression is terminated in all cell types, except bipolar interneurons, which are the last neuronal class to be generated during retinal histogenesis.

Null mutations in Chx10 cause congenital microphthalmia in humans(Ferda Percin et al., 2000)and mice (Burmeister et al.,1996) (previously referred to as ocular retardation oror^J), and antisense RNA injections into zebrafish embryos cause a failure of retinal development(Barabino et al., 1997). Gross abnormalities shared between Chx10-null humans and mice include small eyes, cataracts, iris coloboma and blindness(Ferda Percin et al., 2000;Robb et al., 1978). Although the whole eye is affected by loss of Chx10 function, the primary genetic defect is specific to the retina and is characterized by two major developmental defects: a dramatic reduction in retinal cell number and an absence of bipolar interneurons(Bone-Larson et al., 2000;Burmeister et al., 1996;Konyukhov and Sazhina, 1971). Although it has been suggested that Chx10 acts in combination with the neurogenic bHLH gene Mash1 (Ascl1 — Mouse Genome Informatics) to promote the bipolar cell fate(Hatakeyama et al., 2001), the function of Chx10 in regulating cell number is unknown.

Recent studies have demonstrated the importance of cyclin-dependent kinase inhibitor (CDKI) proteins as negative regulators of proliferation in the developing CNS (Cunningham and Roussel,2001). Two families of mammalian CDKI proteins are known: the Ink4 family, comprising p15^Ink4a, p16^Ink4b,p18^Ink4c and p19^Ink4d; and the Cip/Kip family,comprising p21^Cip1, p27^Kip1 (hereafter referred to as Kip1; Cdkn1b — Mouse Genome Informatics) and p57^Kip2. Ink4 and Cip/Kip proteins are functionally distinct in that Ink4 proteins inhibit Cdk4 and Cdk6, and Cip/Kip proteins inhibit Cdk2, but high levels of expression of any of these proteins is sufficient to block progression through the cell cycle (Nakayama,1998; Sherr and Roberts,1999; Vidal and Koff,2000). Although all of the CDKI genes are expressed in the CNS,only p19^Ink4d, Kip1 and p57^Kip2 have been identified as regulators of proliferation in the retina(Cunningham et al., 2002;Dyer and Cepko, 2000;Levine et al., 2000). Of these three genes, Kip1 appears to be the most important with respect to retinal cell number regulation. Kip1 protein is expressed in most, if not all, retinal cells as they exit the cell cycle during differentiation(Dyer and Cepko, 2001;Levine et al., 2000), and theKip1 knockout retinal phenotype is the most severe, as demonstrated by a high level of ectopic RPC proliferation and focal dysplasia(Cunningham et al., 2002;Dyer and Cepko, 2001;Levine et al., 2000;Nakayama et al., 1996).

Although Kip1 regulates proliferation in a large number of tissues throughout the developing embryo (Fero et al., 1996; Kiyokawa et al.,1996; Nakayama et al.,1996), its activity may be differentially regulated by tissue-specific factors, in order to control the increase in cell number during the proliferative expansion of developing tissues. To investigate this possibility, we sought to determine whether a genetic interaction exists between Chx10 and Kip1 in the developing mouse retina. In this study, we show that Kip1 protein is abnormally present in retinal progenitor cells ofChx10-null mice, and that the genetic elimination of Kip1alleviates the cell number deficit in the Chx10-null retina. Interestingly, lamination is restored in the Chx10, Kip1 double null retina, but bipolar cells are still absent. We further show that Chx10 is not likely to be a repressor of Kip1 gene transcription, and that cyclin D1 (CycD1; CycD1 — Mouse Genome Informatics) may mediate the ability of Chx10 to prevent Kip1 protein accumulation in progenitors.

Chx10, Kip1 double null mice were generated by intercrossingChx10-null and Kip1-null mice. Genotyping of mouse-tail DNA was performed by PCR and subsequent restriction digest to detect mutant and wild-type Chx10 alleles(Burmeister et al., 1996), and by PCR to determine Kip1 mutant and wild-type alleles(Fero et al., 1996).Chx10-null and Kip1-null animals were on a 129/Sv background. Animals were housed in an animal facility and cared for according to IACUC guidelines.

Retinal tissue was obtained by dissecting the surrounding ocular tissues away from the retina in Hanks buffered saline solution (HBSS). For immunohistochemistry, retinas were fixed in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) for 1 hour. The tissue was then cryoprotected in 20% sucrose in PBS, embedded in OCT, and stored at -80°C until sectioning. Sections (12 μm) were used for immunohistochemistry.

The following antibodies were used in this study: rabbit anti-neuronal class III β-tubulin (Covance, Richmond, CA); sheep anti Chx10 (Exalpha Biologicals, Boston, MA); rabbit anti-cellular retinaldehyde binding protein(CRALBP; Dr J. Saari, University of Washington, Seattle); mouse monoclonal anti-CycD1 (Santa Cruz Biotechnology, Santa Cruz, CA); rabbit anti-CycD1/bcl-1(Lab Vision, Fremont, CA); mouse anti-Kip1/p27 (Transduction Laboratories,Lexington, KY); mouse anti-nestin (Developmental Studies Hybridoma Bank, Iowa City, IA); rabbit anti-phospho-Histone H3 (Upstate Biotechnology, Lake Placid,NY); mouse monoclonal anti-proliferation cell nuclear antigen (PCNA Clone PC10; Dako, Denmark); rabbit anti-protein kinase C α (PKCα; Sigma,St. Louis, MO); rabbit anti-recoverin (Dr J. Hurley, University of Washington,Seattle, WA); mouse monoclonal anti-Rhodopsin (Rho 4D2; Dr R. Molday,University of British Columbia); and rabbit anti-calbindin (Chemicon International, Temecula, CA). All antibodies were used at appropriate dilutions in 2% normal goat or donkey serum, 0.15% Triton X-100 and 0.01%sodium azide in PBS. Primary antibodies were followed with species-specific secondary antibodies conjugated to Fluorescein (FITC; Jackson Immunoresearch,West Grove, PA), rhodamine (TRITC, Jackson Immunoresearch), Alexa Fluor 488(Molecular Probes, Eugene, OR) or Alexa Fluor 568 (Molecular Probes). Nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI; Fluka).

To obtain total cell counts for all genotypes at P0, retinas were dissected as above, then trypsinized and subsequently triturated and resuspended in media. Cells were counted on a hemocytometer and the total number of cells per retina calculated. Cells were plated on poly-d-lysine coverslips and allowed to settle for 1 hour at 37°C and 5% CO₂. Cells were then fixed with 4% PFA in PBS and stored at 4°C. Coverslips were stained with antibodies as above.

Quantification of the percentage of cells expressing immunohistochemical markers was done by random field analysis on each coverslip, counting the number of cells positive for that marker and the total number of cells in that field (stained with DAPI). A minimum of 500 cells was counted per coverslip,and the percentage of cells expressing the marker for that coverslip was determined. At least three different animals were analyzed per condition and an average percentage was determined, with each animal counted asn=1. Statistical significance was determined by unpaired t-tests using StatView (Abacus Concepts, Cary, NC).

P0 wild-type tissue was obtained as above for immunohistochemistry. Retinas were fixed on ice for 2 hours in 4% formaldehyde in PBS/2mM EGTA, followed by cryoprotection and storage as above. Sections were cut as above and stored at-80°C until use. In situ hybridization was performed on sections as previously described (Schaeren-Wiemers and Gerfin-Moser, 1993).

Double in situ labeling was performed as for single in situ reactions, with the following modifications: One RNA probe was labeled with dig-UTP and one probe was labeled with Fluorescein-12-uridine-5′-triphosphate (F-UTP,Roche). Hybridization of both probes was carried out at the same time, with each probe at a concentration of 400 ng/ml. After hybridization, sections were incubated with anti-fluorescein-AP (Roche). Probes were visualized by incubating sections with Fast Red tablets (Roche) dissolved in 0.1 M Tris (pH 8.2). After sufficient staining, the reaction was stopped by rinsing in TE (pH 8) and then washing for 10 minutes in 0.1 M Glycine (pH ∼2). The tissue was then incubated with anti-dig-AP as above and visualized using B3 with BCIP/NBT.

Total RNA (1.8 μg) from P0 wild-type and Chx10-null retinas(Trizol, Invitrogen, Carlsbad, CA) was electophoresed in agarose under denaturing conditions and transferred to Nytran SuperCharge membrane(Schleicher and Schuell, Keene, NH). [³²P]-dCTP labeled random primed DNA probes (Ladderman, Takara Biochemical, Berkeley, CA) for Kip1 and CycD1 (see below) were hybridized overnight at 42°C and washed at a final stringency of 0.1×SSC at 72°C (for Kip1) and 90°C (for CycD1). Filters were exposed to Biomax-MS film (Kodak, Rochester, NY) overnight at-80°C with an intensifying screen. Kip1 probe was synthesized from a 590 bp cDNA fragment corresponding to the complete open reading frame (ORF), and CycD1 probe was synthesized from a 250 bp PCR clone that spans 120 nucleotides(nt) of the 5′-untranslated region and 105 nt of the ORF containing the N-terminal domain. Specific details for electrophoresis, transfer and hybridization have been described previously(Chow et al., 1998;Levine and Schechter,1993).

Total retinal cell numbers in Chx10-null mice are much lower than in wild-type controls, and Chx10-null retinas mostly lack the layered organization that characterizes the normal retina(Burmeister et al., 1996) (see below). Because Kip1 is expressed in the developing retina and is known to regulate proliferation negatively in many developing tissues(Chen and Segil, 1999;Dyer and Cepko, 2001;Fero et al., 1996;Kiyokawa et al., 1996;Levine et al., 2000;Lowenheim et al., 1999;Nakayama et al., 1996), we compared Kip1 expression in wild type and the hypoproliferativeChx10-null retinas (Fig. 1). In this and subsequent quantification of protein expression,retinal tissue was dissociated, and cells were plated onto coverslips for immunohistochemical analysis and cell type quantification (see Materials and Methods). We found that a significantly higher percentage of cells were Kip1 positive in Chx10-null retinas than in wild-type retinas (80% versus 49%, P<0.01). This suggests that Chx10 is necessary for normal regulation of Kip1 protein in the developing retina.

To determine if the presence of Kip1 protein contributes to the reduced cell number in the Chx10-null retina, we generated Chx10,Kip1 double null mice. We found that many of the qualitative phenotypic defects of the Chx10-null mouse retina are rescued in the Chx10,Kip1 double null mouse. Differences between a wild-type retina, aChx10-null retina, a Chx10, Kip1 double null retina, and aKip1-null retina at P19 can be seen at the macroscopic level(Fig. 2). Much of the size deficit of the Chx10-null is rescued in the Chx10, Kip1 double null retina. Radial cross-sections of these retinas reveal that the defective lamination of Chx10-null mice is also rescued in the double null retina. In the double null, the outer nuclear layer, inner nuclear layer,ganglion cell layer and the intervening plexiform layers are all easily distinguishable. These results support the hypothesis that Chx10 is necessary to antagonize Kip1 function during retinal development.

To determine if the rescue includes all of the major retinal cell types,retinal sections from P19 mice of each genotype were analyzed with cell type-specific markers. At P19, retinal histogenesis is complete(Sidman, 1961;Young, 1985b). Photoreceptors were visible with recoverin labeling (Fig. 3A-D), and opsin labeling (not shown). Calbindin was used to identify horizontal cells and amacrine cells in the INL, and displaced amacrine and ganglion cells in the GCL(Fig. 3E-H), and Müller glia were seen with CRALBP labeling (Fig. 3I-L). AII and starburst amacrine cells were visible with calretinin labeling (Fig. 3M-P)and ChAT labeling (Fig. 3Q-T),respectively. Although all of these major cell types are present inChx10-null retinas, they do not have normal morphology and are not organized into layers (Fig. 3B,F,J.N,R). By contrast, cells comprising each of these cell types have grossly normal structure and location in the Chx10, Kip1double null retina (Fig. 3C,G,K,O,S).

Mature Chx10-null retinas lack bipolar cells(Burmeister et al., 1996). To determine if bipolar cell development was rescued in the double null, we used markers for PKCα (Fig. 4)and G₀α (not shown) to label rod and ON-cone bipolar cells,respectively. Interestingly, both indicated an absence of bipolar cells inChx10, Kip1 double null mice. These results demonstrate that even when the cell number and size deficit in Chx10-null mice is alleviated, the bipolar cell deficit remains. Therefore, Kip1 is not involved in the role that Chx10 plays in bipolar cell specification or maintenance.

The adult Chx10, Kip1 double null retina is intermediate in size between retinas from Chx10-null and wild-type mice. Part of the difference between wild-type and double null retinas probably arises from the lack of bipolar cells in the double null, as these cells normally constitute∼10% of the total cell number (Young,1985a). We therefore analyzed cell number at P0, a time before the birthday of bipolar cells (Sidman,1961; Young,1985a). Interestingly, DAPI stained sections show that theChx10, Kip1 double null retina has an intermediate number of nuclei at P0 relative to Chx10-null and wild-type retinas(Fig. 5A-C). Calculation of total cell number from dissociated P0 retinas revealed thatChx10-null retinas have 19-fold fewer cells than wild-type retinas,and that Chx10, Kip1 double null retinas are reduced fourfold compared with wild type (Fig. 5D). The proportion of the Chx10-null cell number defect that is independent of bipolar cell genesis is therefore significantly, but not completely, rescued by deletion of Kip1 (see Discussion).

The greatly reduced cell number in Chx10-null retinas could be due to defects in proliferation, increased cell death, premature differentiation or to some combination thereof. Similarly, the rescue of cell number in theChx10, Kip1 double null retina could be due to compensatory changes in any or all of these processes. To begin to dissect the relative contributions of these factors to cell number, we measured the extent of apoptosis, differentiation and proliferation in Chx10-null,Chx10, Kip1 double null and wild-type retinas.

TUNEL labeling was performed at P0 (Fig. 5E-G), P10 and P19 (not shown) to measure apoptosis. We found that there were no significant differences in the tissue distribution of apoptotic cells (Fig. 5E-G), or in the percentage of TUNEL-positive cells between wild-type, Chx10-null andChx10, Kip1 double null retinas(Fig. 5H). This suggests that the cell number rescue in Chx10, Kip1 double null retinas cannot be accounted for by changes in cell death.

If Chx10 and Kip1 regulate cell number by influencing how and when RPCs choose to exit the cell cycle, then the ratios of progenitor cells and differentiated cells should change between genotypes. For example, if the loss of Chx10 function causes premature differentiation similar to that observed due to inactivation of Hes1 (Tomita et al., 1996), then the percentage of progenitor cells in theChx10-null retina should decrease, and the percentage of differentiated cells should increase in comparison with wild type. Furthermore, if the loss of Kip1 compensates for the loss of Chx10 function in the double null retina by delaying differentiation, then the percentage of progenitors should increase at the expense of the differentiated cell population. To address this, we calculated the percentages of retinal cells in mice expressing markers for progenitors (PCNA) and neurons (Class IIIβ-tubulin, recoverin and neurofilament) at P0. Interestingly, the percentages of progenitors and neurons do not change in theChx10-null and Chx10, Kip1 double null retinas when compared with wild type or to each other (Fig. 5I). At this age, progenitors constitute about 50% of cells in all genotypes, despite the large differences in total cell number between genotypes. To verify this, we labeled sections through central retina from each genotype with antibodies for PCNA(Fig. 6A-C). Additionally, we performed in situ hybridization for Chx10 mRNA in sections from each genotype (not shown). Using both methods, we found that the progenitor population size was proportionally maintained across genotypes. TheChx10-null hypocellular phenotype, and the rescue of cell number in the double null, are therefore not attributable to changes in the timing of cell cycle exit, or in the onset of cellular differentiation.

These observations lead to several important conclusions. First, as we see no evidence of changes in the ratios of progenitors, differentiated cells or apoptotic cells across genotypes, the rescued cell number in Chx10,Kip1 double null retinas must be primarily due to an increase in proliferation. Specifically, it is probably due to an increase in the rate of proliferation (i.e. the number of cell cycles over time), as opposed to changes in timing of exit from the cell cycle. If the latter were true, we should have observed an increase in the fraction of differentiated cells, and a decrease in the fraction of progenitors. Second, the rescued proliferation in the double null must be occurring specifically among progenitor cells, as they proportionally maintain their population size in all genotypes. Third,because the proportion of cells that are progenitors in Chx10-null retinas is the same as in wild-type retinas (50%,Fig. 5I), the increase in Kip1 in Chx10-null cells (to 80%, Fig. 1) must be occurring in progenitor cells. Therefore, Chx10 must have some role in preventing Kip1 protein accumulation in retinal progenitors,and the removal of Kip1 from Chx10-null RPCs must be responsible for their increased proliferation in the double null retinas.

Although it has previously been shown that Chx10 is expressed in RPCs(Liu et al., 1994), and that Kip1 is expressed primarily in postmitotic differentiating cells(Levine et al., 2000), these protein patterns have not been directly compared. Double label immunofluorescence with antibodies to Chx10 and Kip1 on the same P0 wild-type mouse section show that these proteins are predominantly expressed in mutually exclusive cell populations (Fig. 7A-C). This result is consistent with the hypothesis that Chx10 negatively regulates expression of Kip1.

As Chx10 is a member of the homeobox family of transcription factors, and because an antagonistic interaction exists between Chx10 and Kip1, we hypothesized that Chx10 acts as a transcriptional repressor of theKip1 gene. If true, then Kip1 mRNA should not be expressed in the same cells as Chx10. Unexpectedly, we found that the mRNAs of both genes are present in the same population of cells: the RPCs(Fig. 7D-H). Furthermore, the steady state levels of Kip1 mRNA are not different in the wild-type and Chx10-null retinas (Fig. 7I). These observations show that Chx10 is not a potent transcriptional repressor of the Kip1 gene, and suggest that Chx10 acts at a post-transcriptional level to prevent Kip1 protein accumulation in RPCs. Our finding that Kip1 protein is present in differentiating cells in which Kip1 mRNA is not present, at least at high levels, strongly suggests that the regulation of Kip1 protein accumulation changes dramatically near the time of cell cycle exit (Fig. 7J) (see Discussion).

Previous studies have demonstrated that CycD1 has a function in promoting RPC proliferation (Fantl et al.,1995; Ma et al.,1998; Sicinski et al.,1995). Interestingly, deletion of the Kip1 gene abrogates this requirement for CycD1 (Geng et al.,2001; Tong and Pollard,2001). It has also been shown that in addition to their canonical role as activators of Cdk4, D-type cyclins can sequester Kip1 away from the CycE:Cdk2 complex in human keratinocytes and mink lung epithelial cells,thereby contributing to progression of cells through G1 phase(Polyak et al., 1994;Reynisdottir et al., 1995;Toyoshima and Hunter, 1994). We therefore hypothesized that in RPCs, CycD1 may mediate negative regulation of Kip1 by Chx10. We first examined the localization of CycD1 protein relative to Chx10 and Kip1 on cross-sections from P0 wild-type retinas. As expected, we found that CycD1 and Chx10 are expressed in the same cells(Fig. 8A-C), while CycD1 and Kip1 are expressed in mutually exclusive cells(Fig. 8D-F). Next, we examined the localization patterns of CycD1 and Kip1 in retinas that were null forChx10 (Fig. 8G-I). We found that CycD1 for the most part retained its complementary expression with Kip1. Finally, we examined the localization patterns of Chx10 and Kip1 in the retinas of P0 CycD1-null mice(Fig. 8J-L,N). In these mice,Chx10 and Kip1 had significant co-expression, and cells that co-expressed the two proteins were in the neuroblast layer. This result contrasts starkly with the mutual exclusion of Chx10 and Kip1 proteins in wild-type retinas, and it supports the hypothesis that CycD1 mediates the antagonism between Chx10 and Kip1 (Fig. 8M).

To quantify these observations, we performed cell counts on dissociated cells that were double labeled with combinations of Kip1, Chx10 and CycD1 antibodies. Kip1 was co-localized with Chx10 or CycD1 in only about 1% of cells in wild-type retinas (Fig. 9A,B). In CycD1-null retinas, however, 24% of cells were positive for both Chx10 and Kip1 (Fig. 9A). In Chx10-null retinas, 8% of cells co-expressed CycD1 and Kip1 (Fig. 9B). These results show that the negative regulation of Kip1 protein levels in RPCs is critically dependent on the presence of both Chx10 and CycD1.

Fig. 9B also shows that the percentage of cells expressing CycD1 is significantly reduced inChx10-null retinas, compared with wild-type (29% versus 48%,P<0.0003). In contrast to Kip1, the change in CycD1 protein expression is correlated with a reduction in steady state levels of theCycD1 mRNA in the Chx10-null retina(Fig. 9C). Interestingly, even though cell number is significantly restored in the Chx10, Kip1double null retina, the percentage of cells expressing CycD1 (30%) is not restored to normal (Fig. 10). Therefore Chx10 appears to be required to maintain the correct expression of CycD1. Together with the evidence that CycD1 is involved in the negative regulation of Kip1, this result establishes a link between Chx10 and both positive and negative regulators of the G1 phase of the cell cycle. These results suggest that the decreased cell number observed in theChx10-null retina is due, in large part, to a change in the balance of CycD1 and Kip1 in retinal progenitors.

The machinery of the cell cycle and the function of proteins such as cyclins and CDKIs are remarkably well conserved in widely divergent species(Fisher, 1997;Follette and O'Farrell, 1997). However, the mechanisms of external and internal regulation of this machinery vary greatly. The Chx10 protein is crucial for regulating cell number in the developing vertebrate retina, but how it accomplishes this is unknown. In this study, we show that the requirement for Chx10 is abrogated in the absence of the G1 phase regulatory protein Kip1, and that the normal regulation of Kip1 by Chx10 involves another G1-phase protein, CycD1. Our analyses of apoptosis and cell type generation in Chx10-null and Chx10, Kip1double null mice indicate that Chx10 does not regulate cell death or the timing of differentiation. Each of these lines of evidence point specifically to a function for Chx10 in driving RPC proliferation by regulating expression of critical cell cycle components. Most importantly, we show that Kip1 aberrantly accumulates in RPCs in the Chx10-null retina, and that this is in large part responsible for the decreased cell number observed in this animal model of microphthalmia.

Our results and earlier studies of Chx10(Barabino et al., 1997;Belecky-Adams et al., 1997;Burmeister et al., 1996;Levine et al., 1997a;Passini et al., 1997) provide important support for the notion that homeobox genes in general might play tissue-specific roles in directing proliferation, especially in the nervous system. The activity of these genes may link tissue-specific regulation of cell number to the internal mechanics and timing of the cell cycle. Other examples of homeobox genes with tissue-specific roles in proliferation include Emx2 and Otx1 in development of cerebral cortex, and Optx2 in Xenopuseyes (Cecchi et al., 2000;Tole et al., 2000;Zuber et al., 1999). If different homeobox genes independently regulate proliferation in different tissues by their influence on cell cycle dynamics, then these genes may have been excellent targets for evolutionary changes of tissue size between vertebrate species.

This study is among the first to demonstrate a genetic interaction between a homeobox gene and regulatory proteins of the cell cycle, and the first to show rescue in an animal model of microphthalmia through deletion of a cell cycle regulator. A study by Bone-Larson et al. demonstrated rescue of theChx10-null phenotype by breeding a Chx10-null allele in 129/Sv mice into a Mus musculus castaneus background. Some mice of this background that were homozygous null for Chx10 had improved retinal size,bipolar cells and measurable ERGs. Although the genetic modifiers that produced this effect are unknown, we believe that a malfunctioning Kip1 protein in this background could not be the sole modifier, as Chx10,Kip1 double null mice do not have bipolar cells. The rescue we observe inChx10, Kip1 double null mice is specific to the proliferation defect,and we believe it reflects the antagonistic actions of these two proteins during development. Unknown genetic modifiers from Kip1-null mice can not account for the rescue we observed, as the phenotype was tightly linked to the Kip1 allele. Our study also shows that Chx10 is still required for bipolar cell genesis or maintenance even when proliferation is restored. Additionally, it shows that Chx10 is not required for the formation of normal retinal lamination and architecture, except to the extent that it is needed to generate the building blocks of the retinal cytoarchitecture through regulation of cell number.

Kip1 is a tumor suppressor that can inhibit the S phase-promoting kinase Cdk2 (Polyak et al., 1994;Toyoshima and Hunter, 1994),and it has been shown to be a key regulator of cell cycle exit in the developing rodent retina (Cunningham et al., 2002; Dyer and Cepko,2001; Levine et al.,2000). Kip1 has been suggested to accumulate in late G2 and early G1 phases of the RPC cell cycle (Dyer and Cepko, 2001), possibly only in the last cell cycle of a cell that will go on to differentiate. We find that in the wild-type retina, Kip1 protein is almost never seen in cells that express Chx10(Fig. 7A-C). Furthermore, Kip1 protein levels are deregulated in the Chx10-null retina(Fig. 1;Fig. 8H-I). The ectopic expression of Kip1 in RPCs does not appear to cause premature cell cycle exit in the Chx10-null retina, as the ratio of progenitors to differentiated cells is normal at P0, a relatively advanced stage of retinal development (Fig. 5I). However,it is clear that Kip1 contributes to the decreased retinal cell number of these mice. The rescued phenotype of the Chx10, Kip1 double null mouse retina demonstrates that in fact a significant amount of the severity of the Chx10-null retina can be attributed to the presence of Kip1. We found that at P0, the Chx10-null retina is 19 times smaller than wild type, while the double null retina is four times smaller than wild type.

An important question is whether the rescue we see in the Chx10,Kip1 double null retina reflects a specific interaction between these two genes among progenitor cells or if, instead, it reflects a nonspecific increase in cell number in another cell population, caused by the removal of the cell cycle inhibitor Kip1. Our findings that Kip1 expression is increased specifically in progenitor cells in the Chx10-null retina, and that progenitor cells contribute proportionally to the increased cell number in theChx10, Kip1 double null retina, demonstrate that the rescue reflects an interaction that is specific to progenitors. Interestingly, in a study examining the relationship between Myc, a positive cell cycle regulatory protein, and Kip1, generation of a double null mouse did not rescue the severe cell number defects of the Myc single null(Trumpp et al., 2001). Their study demonstrates that genetic elimination of Kip1 does not always compensate for the lack of positive cell cycle regulators.

As Kip1 contributes significantly to the phenotype of theChx10-null retina, and because it does not do so by promoting premature cell cycle exit, we propose that its abnormal presence in RPCs is in large part responsible for the reduced proliferation observed inChx10-null mice. An interesting possibility is that increased Kip1 expression lengthens the G1 phase in RPCs, leading to an increase in the total cell cycle time. This suggestion is supported by two independent observations:that transgenic Kip1 mis-expression in cortical progenitors lengthens their G1 phase, and as a consequence, total cell cycle time(Mitsuhashi et al., 2001); and that the cell cycle time in Chx10-null RPCs is longer than in their wild-type counterparts, possibly because of a specific increase in G1(Konyukhov and Sazhina, 1971). More studies are needed to determine whether Kip1 is in fact responsible for longer cell cycle times in Chx10-null retinas. It is somewhat surprising that the abnormal presence of Kip1 in RPCs of Chx10-null mice does not cause premature cell cycle exit, as Kip1 is a component of the cell cycle exit mechanism at the onset of RPC differentiation(Dyer and Cepko, 2001;Levine et al., 2000). It is possible that Kip1 normally works in combination with other intracellular and extracellular factors (Fuhrmann et al.,2000) to promote cell cycle exit, and that its abnormal presence in Chx10-null RPCs has only the effect of slowing the cell cycle.

In other systems, Kip1 has been shown to be regulated transcriptionally,translationally and by at least two post-translational mechanisms(Cheng et al., 1998;Hengst and Reed, 1996;Kolluri et al., 1999;Pagano et al., 1995;Servant et al., 2000). Homeodomain-containing proteins such as Chx10 have long been recognized for their importance in transcriptional control of developmental events, but no transcriptional targets have been identified for the paired-like:CVC homeobox family to which Chx10 belongs. Our data demonstrate that the inhibition of Kip1 by Chx10 cannot be accounted for by a direct transcriptional repression mechanism. First, we find that Kip1 mRNA is present in precisely the same population of cells as both Chx10mRNA and Chx10 protein. Second, Kip1 mRNA levels do not change in theChx10-null compared with wild type. Chx10 must therefore have some post-transcriptional inhibitory effect on Kip1. Furthermore, our results demonstrate that the expression pattern of Kip1 in the wild-type retina is somewhat unusual, with the mRNA and protein seemingly present in complementary cell populations. We suspect that these cell populations are distinct in timing only, with Kip1 mRNA expressed in proliferating RPCs, and Kip1 protein upregulated in the progenitors that exit the cell cycle. This suggests that there must be a significant change in the translational efficiency ofKip1 mRNA or in the stability of the Kip1 protein that occurs around the time of cell cycle exit.

At present, we do not understand the molecular interaction between Chx10 and Kip1. We believe, however, that CycD1 is involved. In addition to activating Cdk4/6, CycD1 can sequester Kip1 away from Cyclin E:Cdk2 complexes,and this may facilitate the targeting of Kip1 for degradation(Hengst and Reed, 1996;Malek et al., 2001;Pagano et al., 1995). The expression patterns we observed suggest that Chx10 may promote CycD1 expression, and consequently prevent Kip1 protein accumulation. We find that the percentage of CycD1 expressing cells is reduced in the Chx10-null mouse retina, and is not restored in the Chx10, Kip1 double null retina, even though cell number is partially restored. This shows that Chx10 is necessary for proper regulation of CycD1 expression. Second, in contrast to wild type retinas, we find that CycD1-null retinas have many cells that co-express Chx10 and Kip1. Over 50% of Chx10-positive cells in theCycD1-null retina were also Kip1 positive (compared with 3% in wild-type retina). This shows that CycD1 is necessary to keep Kip1 protein at low levels or absent in Chx10-positive progenitors. Finally, Kip1 is present in an abnormally high percentage of cells in both CycD1- andChx10-null retinas, suggesting that the activity of both proteins is needed to prevent Kip1 protein from accumulating in cells. Note that it is unlikely that other D-type cyclins are targets of Chx10 function, or inhibitors of Kip1, as neither cyclin D2 nor cyclin D3 is expressed at significant levels in the developing retina(Geng et al., 1999).

Kip1 and CycD1 are known to influence the activity of G1-phase CDKs, and the ectopic expression of Kip1 in Chx10-null retinas is seen in few,if any, cells expressing markers for the S, G2 and M phases of the cell cycle(data not shown). Normal cell cycle progression in retinal progenitors therefore appears to depend heavily on the ability of Chx10, which acts at least in part through CycD1, to prevent Kip1 protein accumulation during G1 phase.

Although our data suggest that Chx10 could be a transcriptional activator of CycD1, it is also possible that its effects on CycD1 may be more indirect. We did not observe Chx10 DNA binding sites in the CycD1 promoter, and CycD1 protein expression does not appear reduced in the cells that express it inChx10-null mice. It may be that Chx10 promotes the expression of other genes required for G1 progression, or that it mediates the transduction of mitogenic signals from the extracellular environment(Anchan and Reh, 1995;Anchan et al., 1991;Chow et al., 1998;Das et al., 2000;Ilia and Jeffery, 1999;Jensen and Wallace, 1997;Levine et al., 1997b;Lillien and Cepko, 1992;McConnell and Kaznowski,1991).

Our data demonstrate that a significant part of the microphthalmic condition of Chx10-null mice (and therefore possibly of humans with mutations in CHX10) is due to the unchecked activity of Kip1 protein in RPCs. However, the intermediate total cell number in the Chx10,Kip1 double null retina compared with wild-type and Chx10-null retinas indicates that the rescue of proliferation is not complete. Furthermore, the hypoproliferative defect of CycD1-null mice at P0 is significantly less severe than that of Chx10-null mice. This evidence demonstrates that Chx10 must act to some extent through mechanisms that are independent of CycD1 and Kip1 to promote proliferation. A complete understanding of the function of Chx10 will require analysis of its potential regulation of other CDKIs, and of other possible roles for Chx10 within the cell cycle.

We thank Mark Hankin, James Roberts, Matthew Fero, Robert Molday, James Hurley and Jack Saari for generously providing mice and reagents; and Sabine Fuhrmann, Richard Dorsky, Monica Vetter and Mahendra Rao for critical comments and suggestions on this manuscript. We thank Mary Scriven and Diana Lim for expert graphics support. We are also grateful for generous funding from Research to Prevent Blindness (Career Development Award to E. M. L.), Prevent Blindness America (Fight for Sight postdoctoral fellowship to E. S. G.), the Foundation Fighting Blindness (Wynn Center for Retinal Degeneration), the University of Utah Research Foundation and the John A. Moran Eye Center.