• Open Access

The role of cell cycle in retinal development: Cyclin-dependent kinase inhibitors co-ordinate cell-cycle inhibition, cell-fate determination and differentiation in the developing retina


  • This work was accepted for inclusion in Developmental Dynamics 238#9 – Special Focus on Visual Systems


The mature retina is formed through multi-step developmental processes, including eye field specification, optic vesicle evagination, and cell-fate determination. Co-ordination of these developmental events with cell-proliferative activity is essential to achieve formation of proper retinal structure and function. In particular, the molecular and cellular dynamics of the final cell cycle significantly influence the identity that a cell acquires, since cell fate is largely determined at the final cell cycle for the production of postmitotic cells. This review summarizes our current understanding of the cellular mechanisms that underlie the co-ordination of cell-cycle and cell-fate determination, and also describes a molecular role of cyclin-dependent kinase inhibitors (CDKIs) as co-ordinators of cell-cycle arrest, cell-fate determination and differentiation. Developmental Dynamics 239:727–736, 2010. © 2010 Wiley-Liss, Inc.


The vertebrate neural retina, comprised of six types of differentiated neurons and one type of glial cells, provides a convenient model for studies of neurogenesis, due to the simple layered structure and easy access for manipulations. During retinogenesis, tight regulation of progenitor-cell proliferation is required for the timed generation of cells in order to achieve proper morphogenesis, histogenesis, and functionality (Dyer and Cepko,2001b; Cremisi et al.,2003; Ohnuma and Harris,2003; Malicki,2004; Donovan and Dyer,2005; Farkas and Huttner,2008; Zhong and Chia,2008). This review briefly summarizes retinal development in the context of cell-cycle regulation and focuses on the multiple roles of cyclin-dependent kinase inhibitors (CDKIs) in the regulation of cell cycle, cell-fate determination, and differentiation in the retina.

Retinal development is initiated with eye-field specification at the anterior part of the embryonic neural tube. Several genes (optx2, rx1, pax6, six3, and tlx) that specify the eye-field have been identified and co-operatively regulate cell proliferation and transcription of various genes required for the initiation of eye formation (Zuber et al.,1999; Casarosa et al.,2003; Miyawaki et al.,2004; Zaghloul and Moody,2007) (Fig. 1A). After induction, the eye field is separated in two eye-primordia that further evaginate to form two optic cups (Fig. 1A and B). During this process, upon the influence of extrinsic growth factors and their downstream components, the retinal precursor cells actively divide to produce sufficient numbers of progenitors (Martins et al.,2008). Cell-cycle activators such as cyclins, cyclin-dependent kinases (CDKs), and proliferating cell nuclear antigen (PCNA) are strongly activated and drive this proliferative stage (Ohnuma et al.,2002; Barton and Levine,2008). Determination factors then decide the fates of retinoblasts, which ultimately differentiate into six major types of retinal neurons (retinal ganglion, amacrine, horizontal, rod, and cone photoreceptor and bipolar cells) and one type of glia (Müller glial cells) (Fig. 1C). The timing of cell-fate determination is strongly associated with down-regulation of the cell cycle and is followed by the differentiation of postmitotic cells into mature neurons.

Figure 1.

Retinal development. The three major developmental processes in Xenopus ocular development are shown. A: Eye-field formation and its separation into two optic primordia. B: Optic vesicle formation and ocular specification. C: Cell-fate determination and cell-cycle regulation in retinal histogenesis.

The tight co-ordination of cell cycle and retinal development is well demonstrated in adult/post-embryonic retinal neurogenesis. The adult fish or amphibian eye has a retinal stem cell population in the most peripheral region of the retina that gradually adds differentiated retinal neurons and glial cells to the central retina throughout the whole life of the animals. This region, known as the ciliary marginal zone (CMZ), has been shown to recapitulate the molecular processes occurring in the developing embryonic retina (Perron et al.,1998; Ohnuma et al.,2002). Figure 2 graphically represents the gene activity and the associated cell-cycle regulation in the CMZ during retinal development. At the peripheral edge of the CMZ, where stem cells reside, low mitotic activity is linked with the transcriptional down-regulation of cell-cycle activators (such as cyclins and CDKs) and the transcriptional activation of eye-field genes such as Rx1 (zone 1) (Ohnuma et al.,2002). Progressing centrally from the peripheral edge of the CMZ (zones 2–3), the cell cycle, triggered by the transcriptional up-regulation of cell-cycle activators, becomes strongly activated to increase the pool of retinal progenitor cells. In the more central region of the CMZ (zone 4), the proliferating activity is down-regulated, while the expression of cell-fate determinants is activated. The cell fate of retinal neurons is largely determined in this central zone and then differentiation to mature neurons occurs (zone 5).

Figure 2.

Structure and cell-cycle progression in CMZ. A: Schematic drawing of Xenopus CMZ. The CMZ is divided into 5 zones based on cell-cycle activity. B: Cell-cycle status in the CMZ.

Higher vertebrates such as mammals and birds are also reported to have retinal stem cell niches in the ciliary body (CB) (Ahmad et al.,2000; Fischer and Reh,2000; Tropepe et al.,2000). The physical location in the eye and the molecular expression pattern of the mammalian CB is highly conserved with that of the CMZ in fish and amphibia. However, it is important to note that their actual role as mammalian retinal stem cells is still under debate (Cicero et al.,2009). Although the cells in the CB have the ability to express retinal cell type markers, they seem to have restricted potential of producing all types of retinal neurons and glial cells. Further comparative analysis among distinct species would be required to elucidate an evolutionally conserved role of this peripheral retinal region.


Numerous studies demonstrate that a majority of developmentally important ligand-induced signaling pathways regulate cell proliferation in the retina. For example, Wnt (Kubo et al.,2003; Denayer et al.,2008), BMP (Murali et al.,2005), TGF-beta (Lillien and Cepko,1992; Anchan and Reh,1995), Shh (Agathocleous et al.,2007; Wall et al.,2009), FGF (Lillien and Cepko,1992; McCabe et al.,1999), and Notch (Ohnuma et al.,2002; Wall et al.,2009) pathways regulate proliferation of progenitor cells or retinal stem cells. These pathways function at all stages through the transition of quiescent retinal stem cells to highly proliferative retinoblasts and postmitotic determined cells. Also, many other intrinsic and extrinsic developmental factors have been reported to regulate cell cycle (Cremisi et al.,2003; Ohnuma and Harris,2003; Martins and Pearson,2008). Although the detailed molecular mechanisms linking to cell-cycle regulation have not yet been determined, increasing evidence suggests that this co-ordination is key in achieving proper retinal development.


In the last decade, a large number of studies have contributed to the advancement of our knowledge on the role of cell-cycle regulation in retinal cell fate determination (Fig. 2, zone 4). Retinal neurons and glial cells are formed in a sequentially conserved order among vertebrates. Retinal ganglion cells are the first cell type to be born in all vertebrates. Amacrine cells, horizontal interneurons, and cone and rod photoreceptors are then generated in a sequence that varies slightly among species, while bipolar cells and Müller glia are the last cell types to differentiate. All differentiated neurons are postmitotic, although Müller glial cells still retain their proliferative potential (Fischer and Reh,2003).

In order to produce postmitotic neurons, the precursor cells need to exit the cell cycle in a timely manner. What is the role of final cell division in neural cell-fate determination? In 1991, an important study was reported by Prof. McConnell's elegant experiments in the developing neocortex (McConnell and Kaznowski,1991). Neural precursors from young animals were labeled with [3H] thymidine and then transplanted at different phases of the cell cycle into an older animal. It was shown that cells transplanted at the S phase of cell cycle were found in the cortical layer 2/3 of the host instead of layer 6, which would have been their natural fate. This indicates that the transplanted cells received determination signals for layer identity during their final cell cycle. The importance of the final cell division in cell-fate specification has been demonstrated by several other findings (see also the following reviews: Cremisi et al.,2003; Ohnuma and Harris,2003; Malicki,2004; Baye and Link,2008; Farkas and Huttner,2008; Zhong and Chia,2008).


(1) The expression of some retinal/neural cell-fate determinants may be activated during the final cell cycle, as in the reported case of Ath5 and Prox1, which are turned on at the G2 phase of retinal progenitors (Dyer et al.,2003; Matter-Sadzinski et al.,2005; Poggi et al.,2005). Also it has been shown that the stability of neural determination factors is being tightly regulated according to the cell-cycle phases (Cisneros et al.,2008). (2) The expression of retinal cell-fate determinants overlaps with the activation of cell-cycle inhibitors (such as CDKIs) and down-regulation of cell-cycle activators (such as cyclins and CDKs) (Ohnuma et al.,2002). (3) Cell-fate determinants can down-regulate the cell cycle and thus influence the timing of the last cell division. In zebrafish, Ath5 is essential for the determination of retinal ganglion cells, the first step in the cell lineage of retinogenesis. Interestingly, deletion of Ath5 skips the timing of cell-cycle exit for and completely disrupts ganglion cell formation (Kay et al.,2001). In other cases, overexpression of proneural genes, such as Xath5, NeuroD, and neurogenin, induces cell-cycle arrest and affects cell-fate determination (Kanekar et al.,1997; Farah et al.,2000; Ochocinska and Hitchcock,2009). (4) It has been demonstrated that forced induction of cell-cycle activation by overexpression of cell-cycle activators inhibits cell-fate determination, whereas cell-cycle inhibition potentiates cell-fate determination in Xenopus retinogenesis and mouse neurogenesis (Ohnuma et al.,2002; Calegari and Huttner,2003). CDKIs have an important role in retinal cell-fate determination in addition to their cell-cycle inhibitory activity (Ohnuma et al.,1999; Carruthers et al.,2003; Vernon et al.,2003). The detailed mechanisms of CDKI activity will be described in the following sections. (5) Recent studies have demonstrated that the dynamics of interkinetic nuclear migration (Fig. 3A) and symmetry of division (Fig. 3B) contribute significantly to retinal/neural cell-fate determination. In the retinal neuroepithelium, the cell bodies of the dividing retinoblasts extend from the apical to the basal side. During the G1, S, and G2 phases of cell cycle, the nuclei dynamically move in the expanded cell bodies in an apico/basal direction, a movement known as interkinetic nuclear migration (Fig. 3A). Towards the M phase, the nuclei migrate to the apical side, divide and then move basally, and then continue this process until their final cell division. Recent analysis by Professor Harris's group challenged this model by showing that except for a small proportion of the apical translocations preceding mitosis, the interkinetic nuclear migration is a rather stochastic event in zebrafish retina (Norden et al.,2009). Using time-lapse imaging, Baye and Link examined the relationship between the location of nuclei at interphases of the cell cycle and the destination of the cells after subsequent divisions (Baye and Link,2007). They found that neuroepithelial cells with greater basal nuclear migrations at the final cell cycle produce postmitotic neurons, while cells with more apical migrations are likely to keep cycling. These observations suggest that during interkinetic nuclear migration, the precursor cells are differentially exposed to extrinsic determination factors that can bias their fate. This idea was further supported by a novel observation in zebrafish mutants (Del Bene et al.,2008). The moks309 mutation in zebrafish causes an increase in the retinal ganglion cells at the expense of bipolar and Müller glial cells. Interestingly, moks309 is a nonsense mutation in the gene of Dynactin-1, which regulates nuclear movement. Del Bene et al. (2008) found that moks309 mutant cell nuclei migrate further to the basal side of the neuroepithelium, where they accumulate. These nuclei are less exposed to Notch activity, which has an apical-high and basal-low gradient. Since Notch activity is known to delay retinal cell-fate determination (Austin et al.,1995; Dorsky et al.,1997; Ohnuma et al.,2002), this lower exposure to Notch activity increases the probability of producing early cell types such as retinal ganglion cells. In accordance with this, the nuclear envelope protein, Syne2a, regulates nuclear position and influences retinal cell-fate determination (Tsujikawa et al.,2007). Another mechanism by which the dynamics of cell division can influence the cell fate refers to the symmetry of division, that is the correlation of the axis and the outcome of the division (Fig. 3B). At the apical side of the neuroepithelium, cells divide largely in two distinct manners. Some cells divide with their mitotic spindle oriented parallel to the plane of neuroepithelium (horizontal division), while other cells divide with their spindles oriented perpendicular to the plane of the neuroepithelium (vertical division). The vertical division is also often called asymmetric division, while the horizontal division is called symmetric division. If cell-fate determinants are distributed non-uniformly, then the asymmetric division may result in the induction of two distinct cell types. In the mammalian retina, such vertical divisions seem to be relatively minor compared to the horizontal ones, although the ratio is developmentally stage-dependent. However, some cells divide with the mitotic spindle orientated at random angles compared to the horizontal or vertical axis. Recent studies have reported that such small variation of division angles can provide significant asymmetric distribution of determinants (Kosodo et al.,2004). This concept of asymmetric division has been extensively studied in lower organisms such as Drosophila melanogaster and Caenorhabditis elegans (Yu et al.,2006; Siller and Doe,2009). In vertebrates, this concept was originally demonstrated by Chenn and McConnell (1995) in cortex development). By live imaging, they showed that horizontal division tends to generate two daughter cells that remain neuroepithelial cells in proliferative status, while vertical division produces differentiating cells. In the rat retina, a horizontal final division tends to produce the same cell type, while a vertical division produces daughter cells with different fates (Cayouette and Raff,2003). In the zebrafish retina, there is no vertical division. However, orientation to the central-peripheral axis creates another asymmetry of the division (circumferential division and radial division) (Das et al.,2003). 3D-time lapse analysis of zebrafish with ath5 promoter-induced expression of GFP showed that circumferential divisions tended to give rise to asymmetric or different fates, while radial divisions tended to produce symmetric or similar fates (Das et al.,2003; Poggi et al.,2005). In Drosophila neurogenesis, many components of asymmetric division that regulate cell polarity have been identified (Yu et al.,2006; Siller and Doe,2009). Also, cell-cycle components, such as cdc2, are known to influence the type of cell division (Tio et al.,2001; Prokopenko and Chia,2005). Recent studies have demonstrated that a similar molecular mechanism regulates asymmetric division of vertebrate retinal/neural progenitor cells (Malicki,2004; Zigman et al.,2005). For example, Zigman et al. (2005) showed that knockout of the mouse inscuteable inhibits horizontal divisions and activates vertical proliferative division in the retina (for more information, see recent reviews: Zhong and Chia,2008; Siller and Doe,2009). In addition, some transcription factors such as Emx2 and Pax6 that regulate cell-fate determination are reported to influence cellular dynamics, although the underlying mechanisms remain to be elucidated (Gotz et al.,1998; Heins et al.,2001,2002; Estivill-Torrus et al.,2002).

Figure 3.

Dynamics of final cell division influence cell-fate determination. A: Interkinetic nuclear migration. During cell cycle of neural precursor cells, nuclei migrate between apical and basal sides depending on the cell-cycle phase. The position of nuclei during interphase and the duration of interphase can influence their cell fates. B: Asymmetric and symmetric divisions. In general, a majority of precursor cells divide by horizontal division, which tends to produce two identical daughter cells that are often still dividing. Thus, it is often called symmetric or proliferative division. During progression of retinogenesis, the ratio of the second type of division, the vertical division increases, producing two distinct cell types. Thus, it is called asymmetric division. Because one daughter cell often differentiates to a neuron, it is also called differentiative division.

Although we described the importance of the final cell division in retinal/neural cell-fate determination, a significant number of studies also indicate that cell-cycle regulation in the final division is not an absolute factor in cell-fate determination. In classical experiments using frog retina, Harris and Hartenstein (1991) showed that pharmacologically induced cell-cycle arrest did not prevent production of all retinal cell types. Also, a recent detailed cell lineage analysis in the cortex indicates that neural cell fates are largely intrinsically determined before the final division (Qian et al.,1998; Shen et al.,2006). It has been shown that, in zebrafish, immature horizontal cells divide to produce more horizontal cells under physiological conditions (Godinho et al.,2007). Moreover, Müller glial cells still retain their proliferative potential after retinogenesis and they can function as retinal stem cells in the chick (Fischer and Reh,2003). Furthermore, some studies showed that FGF signaling does not mediate co-ordination of cell-cycle regulation and FGF-mediated cell-fate determination (McCabe et al.,1999; Martinez-Morales et al.,2005). All these observations indicate that although co-ordination of the final cell cycle is not absolutely required for retinal cell-fate determination, the final cell cycle has a significant contribution to proper retinal cell-fate determination.


As mentioned above, cell-fate determinants regulate cell cycle, while the molecular machinery and cellular dynamics of the cell cycle also influence fate-determination decisions, indicating the importance of a bi-directional co-ordination. However, the detailed molecular mechanisms of co-ordination are still largely unknown. In this section, we will describe a role of CDKIs in the co-ordination of cell-cycle and retinal cell-fate determination. Using the Xenopus embryonic retina system and in order to elucidate the underlying molecular mechanism of cell-fate determination, we identified p27Xic1, a Xenopus member of the Cip/Kip CDKI family, as a co-ordinator of cell-cycle and retinal cell-fate determination (Ohnuma et al.,1999). Three members of the Cip/Kip CDKI family have been identified in vertebrates (Santamaria and Ortega,2006; Besson et al.,2008). In 1993, the first member, p21Cip1, was identified in mammalian systems by four independent groups, followed by identification of p27Kip1 and p57Kip2. Amphibia also have three members, p27Xic1, p16Xic2, and p17Xic3, but they are slightly smaller than the mammalian proteins (Daniels et al.,2004). As indicated in Figure 4A, each CDKI family member has a conserved CDK/cyclin-binding domain at its N-terminal half and member-specific sequences at the C-terminal half (Ohnuma et al.,1999; Daniels et al.,2004). Through direct interaction with a CDK/cyclin complex, CDKIs negatively regulate cell proliferation mainly in G1 phase. Also, p21Cip1 through interaction at the C-terminus with the S-phase activator, PCNA, inhibits S-phase progression.

Figure 4.

Structure of CDKIs and expression of CDKI and NM23 in retina. A: The Kip/Cip CDKI family members have a conserved cyclin/CDK-binding domain at the N-termini and homologue-specific sequences at the C-termini. B: In Xenopus CMZ, p27Xic1 is expressed in the cells undergoing cell-fate determination. The expression of NM23-X3 and X4 starts slightly more peripheral than p27Xic1 but overlaps with p27Xic1 at the more central side. This is probably important for inhibition of early cell-cycle exit and early gliogenesis in retinal lineage.

In amphibia, p27Xic1 is the most highly expressed of the three CDKI members during retinogenesis (Daniels et al.,2004). As previously indicated, amphibian and fish retinas have the CMZ, where neuro/gliogenesis progresses from the periphery to the center throughout the whole lifetime of the animal. Expression of p27Xic1 in the CMZ coincides with the timing of cell-cycle exit in the Xenopus retina. Overexpression of p27Xic1 activates cell-cycle exit, while deletion of p27Xic1 delays cell-cycle exit, indicating that p27Xic1 is a key protein that regulates cell-cycle exit in the retina. This has been supported by studies in zebrafish retina (Kay et al.,2001). In the case of the mammalian retina, the three CDKIs are also expressed (Dyer and Cepko,2001a; Cunningham et al.,2002) and involved in cell-cycle exit. These observations are consistent with the role of CDKIs in neocortical neurogenesis, in which p27Kip1 controls cell-cycle exit from the dividing ventricular epithelium (Tarui et al.,2005).


Our study found that p27Xic1 overexpression in retinal progenitor cells dramatically increases the ratio of Müller glial cells at the expense of bipolar neurons, while its down-regulation decreases the percentage of glia (Ohnuma et al.,1999). The gliogenic activity of p27Xic1 is mediated by its N-terminal half, where the CDK/cyclin-binding domain is located, but is independent of the cell-cycle inhibitory activity. Also, further studies elucidated that through the functional interaction with proneural genes and the Notch pathway, p27Xic1 activates neurogenesis in retinal/neural progenitors (Ohnuma et al.,2002; Vernon et al.,2006). This was supported by the Philpott and Papalopulu groups who found that a p27Xic1 antisense morpholino completely inhibits primary neurogenesis (Carruthers et al.,2003; Vernon et al.,2003). The neurogenic activity of p27Xic1 also depends on its N-terminus in a cell-cycle-independent manner. In a related observation, p57Kip2, a mammalian CDKI, influences amacrine neuron specification (Dyer and Cepko,2001c). These observations indicate that CDKIs regulate both gliogenesis and neurogenesis in a context-dependent manner. Further support for this idea was found in mouse retinal progenitors lacking Math5, a determinant of retinal ganglion cells, which up-regulate p27Kip1, and induce differentiation into Müller glial cells (Le et al.,2006). Similar determination activities of CDKIs have been reported in other developmental events such as muscle and keratinocyte-fate determination (Reynaud et al.,2000; Devgan et al.,2005). In summary, all observations indicate that CDKIs are able to regulate retinal cell-fate determination in addition to their cell-cycle inhibitory activity.

How do CDKIs regulate cell-fate determination? Studies by Philpott's group showed that p27Xic1 activates Xenopus primary neurogenesis by stabilizing the bHLH transcriptional factor, neurogenin (Vernon et al.,2003). In the retina, co-overexpression of p27Xic1 with Ath5, another bHLH protein, potentiates retinal ganglion cell determination (Ohnuma et al.,2002). Also, in the cortex, p27Kip1 interacts with Neurogenin-2 (Nguyen et al.,2006) and this interaction-mediated stabilization influences neural differentiation. Furthermore, p57Kip1 activates myogenesis through direct interaction with MyoD, a muscle-specific bHLH transcription factor (Reynaud et al.,2000), whilst CDKIs directly interact with general transcriptional regulators such as E2F, cMyc, STAT3, and p300/CBP (Delavaine and La Thangue,1999; Coqueret and Gascan,2000; Kitaura et al.,2000; Snowden et al.,2000; Devgan et al.,2005). These observations suggest that such direct interaction with fate-specific transcription regulators is likely to explain CDKI-mediated cell-fate determination in retina.

Although the studies mentioned above suggest that CDKIs can influence cell-fate determination in a manner independent of their cell-cycle inhibitory activity, CDKIs are also likely to indirectly influence cell-fate determination through inhibition of CDK activity because the molecular machinery of the cell cycle and the cellular dynamics during the cell cycle can influence cell-fate determination and are primarily regulated by the activities of CDK and cyclin complexes. To support this idea, Calegari and Huttner (2003) examined the effect of olomoucine, a chemical CDK inhibitor, on the proliferation and differentiation of the telencephalic neuroepithelium. They found that olomoucine treatment induced premature neurogenesis in association with inhibition of proliferation. In addition, several components in developmentally important signaling pathways are regulated by CDK-mediated phosphorylation (Deed et al.,1997; Hara et al.,1997; Zhang et al.,2000; Barnes et al.,2001; Kim et al.,2004; Baughn et al.,2009), suggesting that CDKIs may regulate these processes through regulation of CDK activities.


In addition to their cell-fate determination activity, CDKIs have been reported to regulate cytoskeleton by directly interacting with cytoskeletal regulators through their C-terminal halves (Besson et al.,2008; Frank and Tsai,2009). p21Cip1 regulates actin dynamics through interaction with Rho kinase/ROCK (Lee and Helfman,2004), while p57Kip2 inhibits LIM kinase through their direct interaction-mediated translocation of p57Kip2 to nucleus (Yokoo et al.,2003). Also, p27Kip1 binds directly to RhoA and regulates cell migration (Besson et al.,2004; Nguyen et al.,2006). Actin dynamics are regulated by the Rho pathway, in which ROCK, RhoA, and LIM kinase sequentially function (Fig. 5), and interactions of CDKIs with Rho pathway components potentiate neural differentiation and neural cell migration. Furthermore, RhoA has been reported to influence spindle rotation during cell division of neuroepithelial cells (Roszko et al.,2006), suggesting that CDKIs may contribute to cell-fate determination through regulation of symmetric/asymmetric division. Moreover, p27Kip1 also interacts with Stathmin, a regulator of microtubules (Baldassarre et al.,2005), indicating that CDKIs regulate cytoskeleton by both microtubule and actin pathways. In addition to the role of CDKIs in neural development, cancer cell migration is regulated by cytoskeletal dynamics. Indeed, recent reports show a role of CDKIs as regulators of cancer cell migration (Besson et al.,2008).

Figure 5.

CDKIs co-ordinate cell-cycle arrest, cell-fate determination, and differentiation in neural development. CDKIs are induced by multiple differentiation factors. CDKIs interact with multiple components and regulate three major activities. They bind CDK/cyclin complexes at their N-termini. Also, the N-termini regulate cell-fate determination. On the other hand, the C-termini bind to cytoskeletal regulators. This interaction influences neural differentiation and migration. NM23 also interacts with CDKIs and is likely to influence these three activities of CDKIs.

Although CDKIs have been recognized as inhibitors of the cell cycle, the recent observations mentioned above need a reconsideration of the CDKIs' role. The process of production of postmitotic neurons and glial cells requires cell-cycle arrest, cell-fate determination and differentiation. These processes are triggered by both extrinsic and intrinsic mechanisms and many extrinsic factors such as Wnt (Castelo-Branco et al.,2003), shh (Ohta et al.,2005; Cayuso et al.,2006), TGF-beta (Misumi et al.,2008), BMP (Nakamura et al.,2003), FGF (Li and DiCicco-Bloom,2004), and Notch (Devgan et al.,2005) are known to induce cell-cycle arrest by induction of CDKI expression. As mentioned, CDKIs have two more activities, that of cell-fate determination and differentiation. Interestingly, knockout studies of CDKIs indicate that CDKIs are not essential for cell-cycle progression. These observations clearly demonstrate a key role of CDKIs as coordinators of the three major developmental processes of cell-cycle arrest, cell-fate determination, and differentiation/cell migration in response to their induction by differentiation factors.


CDKIs have major roles in cell-cycle regulation, cell-fate determination, and differentiation/cell migration. In neuronal/retinal cell-fate determination, CDKIs have two opposing activities: activation of neurogenesis and activation of gliogenesis in a context-dependent manner. How do CDKIs co-ordinate these three major activities and how does cellular context influence the choice between neurogenesis and gliogenesis? In Xenopus retina, co-overexpression of p27Xic1 with proneural genes such as Xath5 activates neurogenesis, although the sole overexpression of p27Xic1 activates gliogenesis (Ohnuma et al.,1999,2002). This is also observed in mice (Le et al.,2006), suggesting that the availability of neural determination factors in determining cell fate is important. However, very little is known about the molecular mechanism underlying CDKI-mediated co-ordination of these activities.

Our search for proteins that interact with p27Xic1 has identified NM23-X4 as a binding partner (Mochizuki et al.,2009). NM23-X4 is a Xenopus member of the NM23 family, which is highly conserved from microorganisms to higher vertebrates. In vertebrates, the family consists of nine to ten members (Bilitou et al.,2009). The nomenclature of these proteins includes NM23, a species suffix, and the human ortholog's number, for example, NM23-H2 (human) and NM23-Z3 (zebrafish). We identified 8 Xenopus orthologs of the NM23 family. In Xenopus retinogenesis, NM23-X3 and X4 are highly expressed at the periphery of the CMZ of Xenopus retina and the expression overlaps with p27Xic1 at the central side (Fig. 4B), while other members such as NM23-X1 are not specifically expressed in the CMZ.

All members of the NM23 family directly bind to the N-terminal region of p27Xic1 although NM23-X3 and X4 showed a higher affinity for CDKIs. NM23 family members inhibit p27Xic1-mediated gliogenesis (and potentiate neurogenesis) through their direct protein interaction. Several activities of NM23 family members have been reported, including nucleotide diphosphate kinase (NDPK, which converts ADP or GDP to ATP or GTP), and histidine-dependent protein kinase activities (Wagner and Vu,1995; Hartsough et al.,2002; Srivastava et al.,2006). To elucidate the molecular mechanism, mutants of essential residues (S150 and H148) of NM23-X4 for protein kinase and NDP kinase were constructed. Interestingly, these mutants were shown to lack the ability to inhibit p27Xic1-mediated gliogenesis (Mochizuki et al.,2009). Further, mutational analysis of NM23-X4 and p27Xic1 revealed that the direct binding of NM23 to the N-termini of CDKIs is absolutely required for the inhibition of CDKI's activity. These observations indicate that either of the NM23 kinase activities is important for this function but general attenuation of GTP or ATP level by NDP kinase activity is unlikely to be the mechanism.

In addition to the effect on cell fate, NM23-X4 is able to inhibit p27Xic1-mediated cell-cycle arrest in retinogenesis (Mochizuki et al.,2009). Since the NM23-X4-binding region of CDKIs overlaps with the CDK/cyclin-binding site at the N-termini of CDKIs, the inhibition of cell-cycle inhibitory activity is probably mediated by competition of NM23 binding with CDK/cyclin binding. Since neurogenesis occurs before gliogenesis in the retinal lineage, our study indicates that NM23-X4 helps delay gliogenesis and inhibits precocious gliogenesis in the sequential production of retinal neurons and glial cells through inhibition of glial cell-fate determination and activation of cell cycle.

Another aspect of the interaction between CDKI and NM23 is cancer metastasis suppression. Although NM23-H1 was originally identified as a “cancer metastasis suppressor” (Steeg et al.,1988; Rosengard et al.,1989) and extensive studies demonstrated that NM23-H1 is involved in cell migration of many cancer types (Ouatas et al.,2003), the mechanisms of this suppression have not been fully determined. Since CDKIs have been reported to act as activators of cell migration even in cancer cells (Baldassarre et al.,2005), it is possible that NM23 family members may influence CDKI-mediated cell migration through their interaction with the N-terminal regions of CDKIs.


Over the last decade, many molecular mechanisms of the co-ordination of cell cycle with retinal developmental processes have been elucidated. However, many questions still remain. For example, during eyecup formation, retinal precursor cells actively divide. We need to know how cell-cycle down-regulation is correctly timed in these cells in such a way as to form eyes of the correct size. Cellular dynamics such as asymmetric division and interkinetic nuclear migration influence cell-fate determination. We need to know how the un-differentiated cells know whether to continue to divide or to differentiate to postmitotic neurons. CDKIs influence both neural and glial cell-fate determination. We need to know how precursor cells process CDKIs' activity to create two opposite outcomes. Retinal precursor cells produce all retinal neurons and glial cells in a sequential retinal cell lineage. The progression of the retinal lineage is associated with cell division. In fly neurogenesis, cell division has an important role in the progression of neural lineage (Grosskortenhaus et al.,2005). We need to know whether or not cell division is important for the progression of retinal lineage, and if so, what cell-cycle mechanism functions to ensure the proper progression of retinal cell lineage. In the next decade, molecular mechanisms underlying these aspects will be determined to answer these questions.


This work was supported by a Fight for Sight PhD studentship. We thank Julie Watson and Jacqueline van der Spuy for improving the manuscript. We apologize to colleagues whose work has not been mentioned due to space limitation.