INTRODUCTION

Neural development is a complex process requiring precise coordination between proliferation and differentiation. During gastrulation in Xenopus embryos, a selection of initially undifferentiated, mitotically active cells located within the neural plate withdraw from the cell cycle and generate a defined pattern of primary neurones (Hartenstein, 1989). Cyclin-dependent kinase inhibitors (cdkis) are expressed in many cell types, including cells of the central nervous system (Yan et al., 1997; Durand and Raff, 2000), and may function in their differentiation as well as in cell cycle withdrawal (Dyer and Cepko, 2000; Zezula et al., 2001). Recent experiments in the Xenopus retina have shown that the cdki, p27^Xic1, biases determined neuroblast precursors toward a glial cell fate and this function is distinct from its regulation of the cell cycle (Ohnuma et al., 1999). We wished to determine the role of p27^Xic1 in cell fate determination and differentiation of primary neurones, the first neural cell type to differentiate from the neural plate.

Mammals express three members of the Cip/Kip family of cdkis, p21^Cip1, p27^Kip1 and p57^Kip2 (Sherr and Roberts, 1999). Although highly expressed in many neural tissues, redundancy and inaccessibility has complicated analysis of cdki function in nervous system development in null mouse models (Deng et al., 1995; Nakayama et al., 1996; Yan et al., 1997). As a result, a clear role for cdkis in neural differentiation has not been demonstrated. One major advantage of the Xenopus system is that there is only one described cdki, p27^Xic1, which shows structural and functional characteristics of p21^Cip1, p27^Kip1 and p57^Kip2 (Su et al., 1995; Shou and Dunphy, 1996).

Primary neurone differentiation requires the sequential activation of proneural basic helix-loop-helix (bHLH) transcription factors such as neurogenin (X-NGNR-1) (Ma et al., 1996), Xash3 (Zimmerman et al., 1993) and NeuroD (Lee et al., 1995), the roles of which have been likened to those of MyoD, Myf5 and myogenin during muscle differentiation. Although bHLH genes coordinate cell cycle exit and differentiation in muscle by upregulating cdkis (Guo et al., 1995; Halevy et al., 1995; Parker et al., 1995), no such relationship has been demonstrated in neural tissue.

We present the first clear in vivo evidence that a cdki has an essential role in the decision to adopt a neural fate. We show that p27^Xic1 is highly expressed in cells destined to become primary neurones, and that it is required for primary neurogenesis at a crucial step between X-NGNR-1 and NeuroD. Moreover, p27^Xic1 overexpression promotes ectopic neurone formation, while stabilising the proneural gene, X-NGNR-1.

MATERIALS AND METHODS

Xenopus embryos, fixation and β-galactosidase staining

Xenopus laevis embryos were obtained by standard methods and staged according to Nieuwkoop and Faber (Nieuwkoop and Faber, 1994). Embryo fixation and β-galactosidase staining (200-300 pg injected per embryo) was performed as described (Sive et al., 2000).

mRNA injection and morpholino antisense oligonucleotides

Capped RNAs were synthesised in vitro from nuc-β-gal and p27^Xic1 (Su et al., 1995); p27^Xic1 NT, p27^Xic1 CT, p27^Xic1 35-96 and p21^Cip1 N50S (Ohnuma et al., 1999); X-NGNR-1 (Ma et al., 1996); NeuroD (Lee et al., 1995); p21^Cip1 (Harper et al., 1993); XMyT-1 (Bellefroid et al., 1996); Xath3 (Takebayashi et al., 1997); Xash3 (Zimmerman et al., 1993); and p21^Cip1 NT, p21^Cip1 CT, and p21^Cip1 20-82, using the SP6 Message Machine kit (Ambion). Embryos are acutely sensitive to levels of p27^Xic1 RNA, so each new RNA batch must be carefully titrated for effect. The antisense morpholino oligodeoxynucleotide used was: 5′-GCAGGGCGATGTGGAAAGCAGCCAT-3′ (Gene Tools LLC).

Whole-mount in situ hybridisation, BrdU detection and antibody staining

Whole-mount in situ hybridisation was performed as described (Shimamura et al., 1994). Linearised Bluescript plasmid from X-NGNR-1 (BamH1/T7), NeuroD (Xba1/T7), Nβtub (BamH1/T3) or pCS2 XMyT-1 (BamH1/T7) and p27^Xic1 (BamH1/T7) was used to generate digoxigenin-11-UTP-labeled antisense RNA probes. BM Purple was used as a substrate. Double in situ hybridisation was performed as described (Sive et al., 2000) using BCIP (light blue) and NBT/BCIP (purple) as colour substrates. For 60-70 μm sections, specimens were post-fixed in MEMFA, embedded in 3% low melting point agarose and sectioned by Leica VT1000M vibratome, mounted in 100% glycerol and photographed with Nomarski optics. BrdU analysis was performed as described by Hardcastle and Papalopulu (Hardcastle and Papalopulu, 2000). Whole-mount antibody staining was performed as described previously (Sive et al., 2000) using anti-phospho-histone H3 (TCS Biologicals).

Western blotting

p27^Xic1 was detected using a polyclonal antibody (Ohnuma et al., 1999). Blots were stripped (Chemicon International) and probed with anti-β-tubulin antibody (1:400) (Santa Cruz Biotechnology).

RESULTS

p27^Xic1 expression in the early embryo

p27^Xic1 expression has previously been shown in differentiating muscle and notochord at neurula stages and at lower levels in the neural plate (Ohnuma et al., 1999; Hardcastle and Papalopulu, 2000). To investigate the specific relationship between primary neurogenesis and p27^Xic1, we more carefully analysed p27^Xic1 expression during development using whole-mount in situ hybridisation.

p27^Xic1 is strongly expressed at stage 10.5-11, significantly before neural differentiation, in three stripes lateral to the midline in the prospective neural plate (data not shown), reminiscent of expression of genes involved in primary neurogenesis (Bellefroid et al., 1996; Ma et al., 1996). To examine this resemblance more closely, we have compared p27^Xic1 expression with that of XMyT-1 (Fig. 1) (Bellefroid et al., 1996).

XMyT-1, a zinc-finger transcription factor, is one of the earliest markers of primary neurogenesis. XMyT-1 staining is first detected at stage 11.5 in cells destined to differentiate into lateral, intermediate and medial primary neurones (Fig. 1D) (Bellefroid et al., 1996) and corresponds to the stripes of stronger p27^Xic1 expression (Fig. 1A). Overlapping p27^Xic1 and XMyT-1 staining can be seen at stage 13 by comparing single in situ hybridisations (compare Fig. 1B with 1E) and examining double in situ hybridisations (Fig. 1H). At this stage, p27^Xic1 placodal expression is clear (Fig. 1B, arrows), but the medial neural stripe of p27^Xic1 staining is obscured by stronger stain in the underlying myotome. p27^Xic1 expression is also apparent in the epidermis outside the neural plate between stages 10.5 and 15 (Fig. 1G). By stage 15, p27^Xic1 expression in the medial and intermediate stripes is completely obscured by staining in the underlying myotome. However, p27^Xic1 staining is still clear in the lateral primary neurone stripes (Fig. 1C) and placodal regions (Fig. 1C, arrows), and closely resembles the expression of XMyT-1, X-NGNR-1 and X-Delta-1 (Fig. 1F and data not shown) (Bellefroid et al., 1996; Ma et al., 1996). Moreover, p27^Xic1 expression is clearly seen in the primary neurone precursors in the sensorial layer of the neuroectoderm, both overlying and lateral to the staining in the myotome and notochord (Fig. 1I) (Hardcastle and Papalopulu, 2000). p27^Xic1 is not uniform in the neural plate at stage 15; rather, its scattered distribution is similar to XMyT-1 and is reminiscent of proneural genes whose expression is restricted by Notch-mediated lateral inhibition (Fig. 1C).

In cultured myoblasts, the bHLH factor MyoD can transcriptionally upregulate expression of both muscle structural genes and the cdki, p21^Cip1, thus coordinating cell cycle exit and differentiation (Guo et al., 1995; Halevy et al., 1995; Parker et al., 1995). Analogous bHLH factors such as X-NGNR-1 and Xash3 function during neurogenesis. However, X-NGNR-1, Xash3 and XMyT-1 alone or in combination are unable to upregulate p27^Xic1 expression either in whole embryos or in ectodermal explants (data not shown). Nonetheless, p27^Xic1 is probably regulated in vivo by an unknown bHLH transcription factor or combination of factors, as p27^Xic1 transcription increases on injection of XE12, a bHLH co-factor, and is inhibited in the presence of Id, a bHLH repressor (data not shown) (Lassar et al., 1991; Jen et al., 1992).

Thus, p27^Xic1 is properly spatially and temporally expressed for involvement in primary neurogenesis. Interestingly, p27^Xic1 expression precedes substantial proneural gene expression (Zimmerman et al., 1993; Lee et al., 1995; Ma et al., 1996) but coincides with the time that Rohon-Beard cells (primary sensory neurones) begin to exit the cell cycle (Lamborghini, 1980).

p27^Xic1 is necessary for neural differentiation

To investigate whether p27^Xic1 is necessary for the differentiation of primary neurones, we used antisense morpholino oligonucleotides (Mo) to prevent translation of p27^Xic1 message (Heasman et al., 2000). Injection of p27^Xic1 Mo at the two-cell stage prevents the accumulation of p27^Xic1 protein that normally occurs after the MBT (Shou and Dunphy, 1996), while injection of a control morpholino (Con Mo) has no effect (Fig. 2A).

As predicted, ablation of p27^Xic1 can increase cell proliferation. Embryos injected with 10 ng p27^Xic1 Mo have a 1.27-fold increase in mitotic cells expressing phosphorylated histone H3 (ph3), compared with the uninjected side (n=35, P=0.02, Student's t-test), as well as a small increase in BrdU incorporation (data not shown). The Con Mo had no effect on ph3 or BrdU (data not shown).

Embryos were assayed for primary neurone formation at stage 15 by staining for X-NGNR-1. X-NGNR-1 is the first proneural gene to be expressed in the neural plate, and specifies the formation of primary neurones (Ma et al., 1996). The Con Mo had no effect on any of the neural markers tested (Fig. 2B; data not shown). p27^Xic1 Mo injection had very little effect on expression of X-NGNR-1 at stage 15 (no change in 91% of embryos, n=64), demonstrating that p27^Xic1 is not required for this initial specification event (Fig. 2C). However, p27^Xic1 Mo, but not Con Mo, reduced expression of XMyT-1, which is downstream of, and expressed shortly after X-NGNR-1 (Fig. 2D). NeuroD, a direct downstream target of X-NGNR-1, is also reduced in the absence of p27^Xic1 (Fig. 2E) (Huang et al., 2000), demonstrating a role for p27^Xic1 during early differentiation, but after specification. Interestingly, at stage 18, p27^Xic1 Mo somewhat reduces expression of X-NGNR-1 in 77% of embryos (n=57), indicating that p27^Xic1 may be required for X-NGNR-1 maintenance (data not shown). Terminal differentiation of primary neurones was examined by staining for Nβtub, a marker of differentiated neurones. 96% (n=136) of p27^Xic1 Mo-injected embryos showed significant downregulation or absence of Nβtub on the injected side (Fig. 2F) that persisted until tailbud stages (data not shown).

p27^Xic1 and a mammalian homologue, p21^Cip1, share the ability to induce Müller glial cells at the expense of neural derivatives in the Xenopus retina, indicating that they may play similar roles both in cell cycle regulation and cell fate determination (Ohnuma et al., 1999). Using this functional homology, we investigated whether p21^Cip1 could rescue the loss of primary neurones observed upon p27^Xic1 Mo injection. In one typical experiment, after injection of p27^Xic1 Mo, 55% of embryos (n=29) had no Nβtub-expressing primary neurones on the injected side (Fig. 3A). The remaining 45% of embryos had substantially reduced primary neurone numbers. Neurones in the lateral stripe were counted in the embryos with reduced Nβtub expression, and they displayed an average of 22 neurones on the injected side compared with an average of 45 on the uninjected side (P<0.001). After co-injection with p21^Cip1, 100% of embryos showed some Nβtub-expressing primary neurones (n=29) with an average number of 44 neurones on the injected side compared with 45 on the uninjected side (Fig. 3B) (P<0.001). Thus, depletion of p27^Xic1 protein by Mo injection prevents primary neurone formation. This effect is specific to the loss of cdki function because neurone formation can be rescued by co-injection of the mammalian homologue p21^Cip1.

We wished to determine whether the ability of p21^Cip1 to rescue primary neurogenesis in the absence of p27^Xic1 was solely related to its ability to arrest the cell cycle, or whether p21^Cip1 plays an additional role in determining primary neural cell fate. To address this question we used N- and C-terminal constructs of p21^Cip1 [amino acids 1-89 (p21^Cip1 NT) and 87-164 (p21^Cip1 CT)], which arrest the cell cycle in distinct ways. p21^Cip1 NT blocks cyclin/cdk kinase activity while p21^Cip1 CT binds and inactivates PCNA (Chen et al., 1995). We assayed the ability of these constructs to downregulate proliferation and compared this with their ability to rescue p27^Xic1 Mo-induced primary neurone loss. Injection of 20 pg full-length p21^Cip1 RNA substantially downregulates BrdU incorporation (data not shown) and rescues primary neurones (Fig. 3B). Although similar amounts of NT and CT construct RNAs (10 and 50 pg, respectively) greatly reduce or eliminate BrdU incorporation at this stage (data not shown), only expression of p21^Cip1 NT rescues primary neurogenesis (68% of embryos, n=38) (compare Fig. 3C with 3D). These data demonstrate that cell cycle arrest alone is insufficient to promote primary neurone formation, but rather some further function of the N terminus of p21^Cip1 is required.

Ohnuma et al. (Ohnuma et al., 1999) have shown that a mutant of p27^Xic1 (amino acids 35-96) blocks overall cdk2 kinase activity but cannot induce Müller glial cells, indicating a distinct, non-cell-cycle requirement for the p27^Xic1 NT region upstream of amino acid 35. To investigate whether a homologous deletion of p21^Cip1 can rescue primary neurones, we co-injected the mutant p21^Cip1 20-82 with p27^Xic1 Mo. Although 50 pg of p21^Cip1 20-82 was able to arrest the cell cycle by inhibiting overall cdk2 kinase (data not shown), it did not rescue primary neurogenesis (Fig. 3E). Thus, a separable differentiation function of p21^Cip1 is required to induce primary neurones.

To determine whether inhibition of cdk kinase activity is necessary for rescue, we injected the p21^Cip1 N50S point mutant, previously shown to have a reduced capacity for cdk2 inhibition (Ohnuma et al., 1999). p21^Cip1 N50S rescued the loss of primary neurones caused by injection of p27^Xic1 Mo in 66% of embryos (n=45, Fig. 3F). These data again show that the ability to promote primary neurogenesis is separable from the ability to inhibit overall cdk2 kinase activity.

p27^Xic1 is required between X-NGNR-1 and NeuroD to promote neural fate determination

When overexpressed, X-NGNR-1 induces extensive ectopic neurogenesis on the flank of the embryo (Ma et al., 1996). Our in situ studies show that p27^Xic1 is expressed in the epidermis outside the neural plate at late gastrula and early neural plate stages (Fig. 1G) and is therefore available to cooperate with ectopic X-NGNR-1 to promote neurogenesis both inside and outside the neural plate. Thus, we investigated whether p27^Xic1 is required for X-NGNR-1-mediated primary neurone formation in the neural plate and embryonic skin. Embryos were injected with X-NGNR-1 and either Con Mo or p27^Xic1 Mo. While embryos co-injected with X-NGNR-1 and Con Mo show extensive ectopic neurogenesis in the epidermis (95%, n=37) (Fig. 4A), both endogenous and ectopic primary neurogenesis is substantially reduced or eliminated in the presence of p27^Xic1 Mo (100%, n=36) (Fig. 4B). The downregulation of Nβtub by p27^Xic1 Mo is rescued by co-injection of p21^Cip1 (97%, n=40) (Fig. 4C), again indicating that the effect of p27^Xic1 Mo on neurogenesis is specific to loss of cdki activity.

NeuroD, a bHLH factor that promotes terminal neural differentiation, lies downstream of X-NGNR-1 (Lee et al., 1995). To determine whether p27^Xic1 acts up- or downstream of NeuroD, we investigated Nβtub expression in response to NeuroD in the presence of Con Mo or p27^Xic1 Mo. As expected, NeuroD induced extensive ectopic neurogenesis in the presence of Con Mo (Fig. 4D). Strikingly, however, although p27^Xic1 Mo almost completely blocked Nβtub expression in 94% (n=70) of X-NGNR-1-injected embryos, NeuroD induced ectopic Nβtub in 82% (n=120) of embryos even in the presence of p27^Xic1 Mo (Fig. 4E). Therefore, p27^Xic1 is required at a crucial early step in neurogenesis and acts between X-NGNR-1 and NeuroD.

p27^Xic1 promotes primary neurogenesis

We have shown that p27^Xic1 is necessary for primary neural differentiation. Next, we wished to determine whether p27^Xic1 expression is instructive for primary neurone cell fate determination and differentiation. Injection of high doses of p27^Xic1 message (250 pg and above) results in rapid cell cycle arrest and embryonic death, triggered by extensive apoptosis at early gastrula stages (Hardcastle and Papalopulu, 2000; Finkielstein et al., 2001). However, if lower doses are injected, cells divide at a slower rate but apoptosis is greatly reduced or eliminated. After careful titration experiments, we chose to inject 30-60 pg p27^Xic1 RNA, a dose that gives modest overexpression compared with endogenous levels (Fig. 5G). Injecting 30-60 pg of p27^Xic1 into one cell of two-cell stage embryos causes, on average, 7% embryonic death. The surviving embryos appear healthy, although slowed cell division is evidenced by a 61% decrease in ph3 expression (n=10, P<0.001, Student's t-test) and a reduction of BrdU incorporation (81% of embryos, n=16) on the injected as compared with the uninjected side (Fig. 5A).

p27^Xic1 overexpression upregulates the number of primary neurones on the injected side in the majority of embryos (67%, n=147) (Fig. 5B,C), but only within the primary neurone stripes in the neural plate. p27^Xic1 does not lead to an appreciably earlier expression of Nβtub and, therefore, is unlikely to cause early birth of primary neurones (data not shown). Interestingly, ectopic p27^Xic1 appears to induce neurones in the superficial layer of the neuroectoderm (Fig. 5C), which is usually refractory to primary neurogenesis (Chalmers et al., 2002). Expansion of primary neurogenesis in the neural plate is clear even in embryos where p27^Xic1 RNA is overexpressed only in the ectoderm, as determined by β-gal tracer staining (Fig. 5C). Therefore, the ectopic induction of primary neurones by p27^Xic1 is cell autonomous and not a secondary result of effects on the underlying myotome (Vernon and Philpott, 2003). To exclude the possibility that an increase in primary neurone cell size was responsible for the apparent increase in Nβtub expression, we used magnified images to count the number of Nβtub-positive cells in the lateral stripe on the injected versus the uninjected side. Of the 67% of embryos that showed an increase in the number of primary neurones, the injected side had, on average, 1.9 times as many neurones as the uninjected side (P=0.02, Student's t-test). This effect is dose dependent; at 30 pg, 39% of embryos had increased Nβtub expression, while 65% of embryos upregulated Nβtub at 60 pg.

To determine which region of p27^Xic1 is required to promote ectopic primary neurones, we injected p27^Xic1 NT (1-96), p27^Xic1 CT (97-210) or p27^Xic1 35-96. While the NT induced ectopic primary neurones (Fig. 4D) (n=123, average 1.7 fold increase, P<0.001), neither the CT nor 35-96 had any effect (Fig. 5E,F), again indicating that p27^Xic1 has a separable N-terminal function required to promote primary neurogenesis that is distinct from its ability to inhibit overall cdk2 kinase activity or block the cell cycle.

In cultured muscle cell systems, the half-life of MyoD protein is extended by co-expression of p57^Kip2 (Reynaud et al., 1999). Originally, this stabilising property of p57^Kip2 was attributed to its inhibition of cdk-dependent phosphorylation of MyoD (Reynaud et al., 1999). However, a non-phosphorylatable form of MyoD (MyoD^Ala200) is also stabilised by co-expression of p57^Kip2 (Reynaud et al., 2000) and an N-terminal α-helix domain of p57^Kip2 is implicated in direct binding with the basic domain of MyoD and masking of potential degradation signals (Abu Hatoum et al., 1998). Because the processes of myogenesis and neurogenesis are highly analogous, we investigated the possibility that p27^Xic1 was promoting neurogenesis by stabilising X-NGNR-1 protein. We injected 50 pg of Myc-tagged X-NGNR-1 alone or in combination with 45 pg p27^Xic1, 30 pg NT, 50 pg CT or 50 pg 35-96 and performed western blots on stage 20 embryos. Co-injection of X-NGNR-1 with p27^Xic1 and p27^Xic1 NT significantly increased the amount of X-NGNR-1 protein persisting at this stage, while the CT and 35-96 had no effect (Fig. 6). However, neither native nor overexpressed p27^Xic1 immunoprecipitates with Myc-tagged X-NGNR-1 from embryo extracts (data not shown), indicating that p27^Xic1 may not stabilise X-NGNR-1 by direct binding. Therefore, p27^Xic1 both stabilises X-NGNR-1 protein and promotes neurogenesis in a manner that is distinct from its ability to inhibit cdk2 activity or arrest the cell cycle, but that may not involve direct binding. By analogy with MyoD, such mechanisms could involve regulation of DNA binding or nuclear localisation (Abu Hatoum et al., 1998; Floyd et al., 2001), but these possibilities remain to be investigated.

DISCUSSION

p27^Xic1 is highly expressed in cells of the neural plate destined to form primary neurones (Fig. 1). Overexpression of p27^Xic1 promotes the formation of primary neurones (Fig. 5), while depletion of p27^Xic1 protein prevents neural differentiation (Fig. 2). Thus, p27^Xic1 is crucial for the specification and differentiation of neurones during early embryogenesis, and has a role distinct from its ability to inhibit overall cdk2 kinase activity or arrest the cell cycle. The discovery that p27^Xic1 is required downstream of X-NGNR-1 but upstream of NeuroD, suggests that this function occurs at the time of neural fate commitment, rather than during the final stages of neural differentiation. Indeed, terminal differentiation of primary neurones driven by NeuroD can occur in the absence of p27^Xic1 (Fig. 4). Furthermore, p27^Xic1 may promote primary neurogenesis via its ability to stabilise X-NGNR-1 protein (Fig. 6).

In the retina, p27^Xic1 regulates neural versus glial cell fate choice in addition to its role in cell cycle inhibition (Ohnuma et al., 1999). Cdkis have been previously implicated in the differentiation of neural derivatives in cultured cell systems such as oligodendrocytes and PC12 cells (Durand and Raff, 2000; Erhardt and Pittman, 1998). However, although highly expressed in many neural tissues, redundancy has complicated analysis of the roles of cdkis during development of the nervous system in null mouse models (Deng et al., 1995; Nakayama et al., 1996; Yan et al., 1997). Nevertheless, several lines of evidence indicate that mammalian cdkis may have analogous roles to p27^Xic1 during neurogenesis and gliogenesis. First, in the PC12 neural cell line, p21^Cip1 overexpression enhances differentiation in addition to arresting the cell cycle (Erhardt and Pittman, 1998). Secondly, mice null for p27^Kip1 have disrupted retinal cell differentiation and organisation, processes regulated by proneural gene function (Nakayama et al., 1996). Moreover, loss of p57^Kip2 causes inappropriate S-phase entry in the retina. Importantly, p57^Kip2 levels drop on cell cycle exit but it is re-expressed postmitotically in a subset of amacrine neural precursor, and loss of p57^Kip2 leads to a significant increase in the calbindin amacrine cell subtype (Dyer and Cepko, 2000). These data indicate that p57^Kip2 is required for proper amacrine cell subpopulation distribution in vivo a process also regulated by proneural genes. Furthermore, Zezula et al. (Zezula et al., 2000) have demonstrated that p21^Cip1-/- oligodendrocytes undergo prompt cell cycle withdrawal, but fail to differentiate. These results indicate that, in some contexts, cdkis may play a more general role in neural cell fate determination and differentiation, in addition to their ability to arrest the cell cycle. However, in contrast to the role of p27^Xic1 during primary neurone differentiation, Ohnuma et al. (Ohnuma et al., 2002) show that cdk kinase inhibition is essential to the ability of p27^Xic1 to synergise with proneural genes and promote early neural cell types in the retina.

The data presented here indicate that cell cycle exit is essential but insufficient for primary neurone differentiation. The NT and CT of p27^Xic1 both arrest the cell cycle (data not shown), but only the NT induces ectopic primary neurones (Fig. 5). Indeed, previous studies have shown that premature cell cycle arrest does not affect primary neurogenesis (Harris and Hartenstein, 1991; Kroll et al., 1998). Instead, our results suggest that a differentiation function of p27^Xic1, which is complementary to, but distinct from, its ability to arrest the cell cycle is required for primary neurone formation. An N-terminal mutant of p27^Xic1 that retains its ability to inhibit cdk2 kinase activity but is missing amino acids 31 to 35 is unable to induce extra primary neurones (Fig. 5F). Additionally, only the full NT of p21^Cip1 can rescue the p27^Xic1 Mo-induced loss of primary neurones (Fig. 3). Moreover, the p21^Cip1 N50S mutant, which has an impaired ability to inhibit cdk2 but contains an otherwise intact N-terminal region, is able to rescue the p27^Xic1 Mo phenotype (Fig. 3F).

Overexpression of p27^Xic1 results in the formation of ectopic neurones, but only within the neural plate (Fig. 5). p27^Xic1 does not lead to an appreciable earlier expression of Nβtub and therefore is unlikely to cause early birth (data not shown). p27^Xic1 doses higher than 75 pg did not enhance, and sometimes inhibited, primary neurone formation (data not shown), indicating acute sensitivity to p27^Xic1 levels. This sensitivity may account for the difference between results presented here and those of Hardcastle and Papalopulu (Hardcastle and Papalopulu, 2000) who failed to detect ectopic neurones inside or outside of the neural plate on injecting 250pg of p27^Xic1 message, a dose that induced 70% embryonic death. High doses of p27^Xic1 may arrest neural plate cells before instructive proneural genes have accumulated or may result in selective death of the overexpressing cells. Indeed, work in keratinocytes has indicated that sustained overexpression of p21^Cip1 at high levels may, in fact, inhibit differentiation (Di Cunto et al., 1998). However, our data clearly indicate that modest p27^Xic1 overexpression promotes primary neurone formation within expanded proneural domains.

Full-length p27^Xic1 and p27^Xic1 NT stabilise X-NGNR-1 protein (Fig. 6) and promote primary neurogenesis (Fig. 5), although the molecular mechanism of this stabilisation is unknown. Initial co-immunoprecipitation studies indicate that p27^Xic1 does not bind directly to overexpressed X-NGNR-1 (data not shown). Future investigations into the effects of p27^Xic1 on X-NGNR-1 DNA binding, nuclear localisation and degradation may reveal alternative mechanisms for the observed upregulation of primary neurogenesis upon p27^Xic1 overexpression.

In summary, the model presented in Fig. 7 proposes that p27^Xic1 acts in parallel with X-NGNR-1 at an early stage of neurogenesis, and requires functions independent of its ability to arrest the cell cycle or inhibit overall cdk2 kinase activity. p27^Xic1 may function in part by regulating X-NGNR-1 stability, and this is also independent of its ability to inhibit cdk2. By combining distinct cell cycle and differentiation functions, a single molecule, p27^Xic1, may provide a powerful way to coordinate the processes of division and differentiation during primary neurogenesis.