Dual Control of Neurogenesis by PC3 through Cell Cycle Inhibition and Induction of Math1

Generation of inducible PC3 transgenic mice

To test the hypothesis that PC3 plays a critical role in the differentiation of neuronal progenitors, we produced mice carrying a PC3 transgene. We chose the conditionally regulated binary tet-on/off system for its versatile spatiotemporal control of gene expression through a tetracycline-responsive tissue-specific promoter (Kistner et al., 1996). The first transgenic mice we generated, Tg TRE-PC3 (Fig. 1A), carried the coding region of the rat PC3 cDNA, conditionally able to be activated under the control of the tetracycline-responsive element (referred to as TRE here, or as TetO by Kistner et al., 1996).

To activate the expression of Tg TRE-PC3 required breeding with a second transgenic mouse carrying the tTA, under the control of a promoter of choice (see also Materials and Methods). We chose to drive the expression of the exogenous PC3 from early developmental stages, before the activation of endogenous PC3 expression, and preferentially, but not exclusively, in neuronal tissues, similarly to the expression of endogenous PC3. We thus generated a second transgenic line, Tg βACT-tTA, carrying the tTA coding sequence under the control of a β-actin promoter that satisfied the requirement of early and mainly neuronal expression (Fig. 1B). As a complementary approach to test our hypothesis, we also targeted the expression of PC3 specifically to neuroepithelial cells using a previously described transgenic line expressing a tetracycline-regulated transactivator under the control of the nestin promoter, which is active in neuronal progenitors (Tg nestin-rtTA) (Mitsuhashi et al., 2001).

Thus, four Tg TRE-PC3 lines (A, C, G, and L) were generated (Fig. 1A). As judged by viability, phenotype, body weight, and number of progeny, these lineages did not differ from wild type (Table 1, and data not shown). Therefore, we excluded alterations arising from either possible toxic effects of the TRE-PC3 sequence per se or its genomic insertion loci. Southern blot analysis, performed using different fragments of the transgene as probes, indicated that the transgene element was inserted in families A, C, G, and L of Tg TRE-PC3, in tandem repeats of 6, 3, 5, and 10 copies, respectively (Fig. 1C,D). Flanking integration fragments were identified and appeared clearly unique for TRE-PC3 A and G, indicating a unique integration site of the transgene (shown with the structural organization in Fig. 1D). The transgene appeared stable, as assessed by Southern blot analysis, routinely performed throughout new generations.

Eight Tg βACT-tTA mice lines were also generated, carrying a transgene in which the human β-actin promoter drives the expression of the tTA sequence (Fig. 1B), and these were analyzed by crossing the Tg βACT-tTA strains with a luciferase standard responder (Tg TRE-Luc, named L7 indicator) (Kistner et al., 1996). Luciferase activity was measured in homogenates of different organs from the bitransgenic mice. Mice with no transgene and of the L7 strain were used as negative controls, whereas a positive control was obtained by breeding the CMV-rtTA strain (Kistner et al., 1996) with the L7 indicator (binary Tg CMV-rtTA/TRE-Luc). We identified a Tg βACT-tTA mice lineage (named nACT37) in which the β-actin promoter was active mainly in the brain, cerebellum, and eye, from the early stages of development to adulthood (A. Servadio, unpublished data). In fact, the levels of luciferase activity in brain, cerebellum, and eye were ∼1000-, 40-, and 200-fold higher, respectively, than in the L7 control mice tissues (Fig. 1E). A strain with no increase of luciferase activity above L7 control mice was also identified (nACT75) (Fig. 1E). No toxic side effects of tTA in the Tg βACT-tTA lineage nACT37 were detected by analyzing viability, phenotype, and body weight (Table 1 and data not shown).

PC3 binary transgenic mice display lethality, dwarfism, and altered neurogenesis

To induce the expression of PC3, lineages A, C, G, and L of the Tg TRE-PC3 were crossed with the Tg βACT-tTA (lineage nACT37). In the absence of doxycycline (a tetracycline analog), the bitransgenic βACT-tTA/TRE-PC3 animals produced the transcription transactivator tTA, which could bind the TRE element within TRE-PC3 to trigger expression of the exogenous PC3. Conversely, doxycycline inactivated tTA and thus repressed expression of the exogenous PC3. We monitored the phenotype, viability, body weight, and number of the progeny after continuous expression of the exogenous PC3 (Fig. 2A–E; Table 1; and data not shown) and analyzed its expression pattern by RT-PCR in the offspring and by whole-mount ISH in embryos (Fig. 3A–D). When mice were kept continuously under doxycycline treatment, the progeny was normal (Table 1).

In contrast, the progeny of bitransgenic mice of lineages A and G, which were never exposed to doxycycline (thus with active transcription of transgenic PC3 since conception), presented a significant percentage of death at birth and a reduced size. The surviving mice usually did not reach adulthood, but, when this occurred, a slower growth rate was observed, associated with abnormal posture and gait and an incomplete development of the cerebellum (Table 1; Fig. 2A–C). The cerebellar folia were less deep and the lobules were of smaller size and length, particularly in the vermis (lobules VIII and IX) and rostrally (lobules IV and V) (Fig. 2D,E). Moreover, the reduction in size was more pronounced in the cerebellum than in the whole brain (55 vs 30%) (Fig. 2D,E). The motility of the surviving Tg βACT-tTA/TRE-PC3 mice was markedly reduced, accompanied by ataxic gait and extension of the lower limbs as early as P10. The frequent appearance of reduced size and weight in bitransgenic βACT-tTA/TRE-PC3 families G and A (Table 1; Fig. 2C) was consistent with the high copy number and the unique integration site of TRE-PC3 elements in the genome. Thus, families A and G of Tg TRE-PC3 were chosen for additional study and considered equivalent.

High expression of exogenous PC3, comparable with that of the endogenous PC3, was observed by RT-PCR on E13 embryos (head) of bitransgenic mice lineages A and G (Fig. 3A,B). Whole-mount ISH performed on E12 bitransgenic embryos indicated that exogenous PC3 transcription was under the control of doxycycline, mainly in the telencephalon, rhomboencephalon, retinal primordium, spinal cord, and dorsal root ganglia, but also in paws and to a lower extent in viscera (Fig. 3D). No evidence of altered morphology of the embryo at E12 could be observed (Fig. 3D, left). This spatial distribution indicated that, in Tg βACT-tTA/TRE-PC3, the expression pattern of the exogenous PC3 in the CNS was similar to that of the endogenous PC3 (Iacopetti et al., 1994). Furthermore, when the expression of exogenous PC3 was activated ∼1 week after birth (by removing doxycycline at birth, given the lag for metabolization of maternal doxycycline), the progeny was normal (data not shown). In these animals, at P40, exogenous PC3 expression was detected by RT-PCR analysis, with a pattern similar to that detected previously by luciferase assay in the Tg βACT-tTA lineage nACT37 (high expression in brain and cerebellum and low or no expression in muscle) (Fig. 3C and data not shown). As a whole, the above results suggested that the expression of exogenous PC3 affected the viability and embryonic development acting after E12 and possibly until the first postnatal week, which raised the possibility of a negative selection of embryos that highly expressed transgenic PC3.

As a complementary approach, we examined the effect of targeting PC3 to neuroepithelial cells, crossing the Tg TRE-PC3 lineages A and G with Tg nestin-rtTA (Mitsuhashi et al., 2001). In this mouse model, doxycycline stimulates a modified tTA transactivator (rtTA), causing the activation of the transgene in binary mice (Kistner et al., 1996). Whole-mount ISH performed in E12 bitransgenic embryos indicated that doxycycline-induced transgenic PC3 mRNA was expressed almost exclusively in proliferating neuroepithelia of CNS and PNS (Fig. 3D, right panels) (Mitsuhashi et al., 2001). No overt alteration of phenotype was observed in newborn and adult bitransgenic mice. However, CNS analysis revealed a number of alterations in the development of the neural tube and of the cerebellum, which are described below.

Expression of PC3 in the neural tube inhibits cell cycle and increases differentiation of neuronal progenitors

In Tg nestin-rtTA/TRE-PC3, a strong expression of the doxycycline-induced exogenous PC3 was detected at E12.5 in the neural tube and dorsal root ganglia and to a lesser extent in the cranial nerve ganglia, in which the endogenous PC3 is normally present (Iacopetti et al., 1994) (Fig. 4A,F). To verify whether overexpression of PC3 affected neurogenesis, we examined differentiation of the neuronal progenitors in these regions. Strikingly, PC3 overexpression led to an increment of βIII tubulin and MAP-2-positive cells (Fig. 4B,C,G,H). This was particularly evident in the neural tube, in which most of the neuroblasts were highly positive for both PC3 and βIII tubulin (Fig. 4B,G,K,L). Concomitantly, a marked decrease of BrdU incorporation was evident in the VZ of the neural tube of the activated Tg (Fig. 4D,I), indicating a decreased proliferation of neuroepithelial cells. Quantitative analysis of the ratio of βIII tubulin⁺ cells/total number of cells in the neural tube revealed an increase of approximately twofold compared with control mice (Fig. 4M). Moreover, the ratio of neural tube cells positive for bothβIII tubulin and PC3/total number of cells (a measure that correlated the expression of βIII tubulin to that of PC3) increased also by twofold, indicating an enhancement of neurogenesis in correlation to the expression of PC3 in the neural tube (Fig. 4M).

Analysis of apoptotic cells in adjacent sections of the neural tube, as detected by TUNEL assay, revealed a nonsignificant increase (from 9.7 ± 1.6% in control Tg to 12.0 ± 1.9% in activated Tg) (Fig. 4E,J,M). Consistently, the number of cells per area, as well as the total area of the neural tube, did not change significantly (Fig. 4M). The fact that these latter two parameters remained constant indicated that the total number of βIII tubulin⁺ neurons increased. In Tg nestin-rtTA/TRE-PC3, at E14, a strong increase of βIII tubulin⁺ newborn neurons was also observed throughout the neural tube (hindbrain, midbrain, and forebrain regions) concomitantly with PC3 overexpression (data not shown).

Similarly, the Tg βACT-tTA/TRE-PC3 was examined at E9.5, a stage of development at which the levels of endogenous PC3 are very low in the neural tube. Analysis of transverse sections of the neural tube at the cervical level (Kaufman, 1999, corresponding to his plate 19a, section i–k) indicated that, in the presence of exogenous PC3 expressed in the ventricular and mantle zones, the percentage of βIII tubulin⁺/total number of cells doubled (from 21.1 ± 1.3% in control Tg to 40.9 ± 3.8% in activated Tg), that apoptosis increased (from 2.7 ± 0.3% in control Tg to 4.5 ± 0.8% in activated Tg), and that the cell number per area and total area did not present significant changes (20.7 ± 1.7 cells/1000 μm² and 1.5 ± 0.1 μm² × 10⁵ in control Tg vs 18.6 ± 1.5 cells/1000 μm² and 1.5 ± 0.1 μm² × 10⁵ in activated Tg).

Together, these data indicate that overexpression of PC3 in the neural tube inhibited cell cycle progression and increased the generation of newborn neurons.

PC3 overexpression in the cerebellum arrests proliferation, increases differentiation of GCPs, and induces Math1 expression

The altered cerebellar phenotype observed at P30 in Tg βACT-tTA/TRE-PC3 (Fig. 2D,E) prompted us to examine the developmental stages in which active proliferation of GCPs takes place.

We first analyzed cerebella at P1. At this stage, the decreased length of lobules and reduced foliation are already evident in both Tgs βACT-tTA/TRE-PC3 and nestin-rtTA/TRE-PC3 (Fig. 5A,G,M,S).

Expression of endogenous PC3 mRNA in the control animal was detectable in the outer EGL, in the molecular and in the Purkinje cell layers and to a lower extent in the IGL (Fig. 5H,T). In contrast, expression of the endogenous PC3 protein was more evident in the outer part of the EGL, decreased in the molecular and the Purkinje cell layers, and was barely detectable in the IGL (Fig. 5J,V). The difference between the expression profiles of PC3 mRNA and protein suggests a translational or posttranslational control. As a whole, PC3 protein expression was less pronounced in regions in which postmitotic, premigratory, or postmigratory granule cells were present. This was similar to observations in the neural tube, in which the PC3 protein was expressed mainly in the VZ and to a lower extent in the surrounding postmitotic mantle zone (Iacopetti et al., 1994, 1999).

Expression of the exogenous PC3 mRNA and protein in cerebellum of the Tg nestin-rtTA/TRE-PC3 matched the spatial and temporal expression pattern of the endogenous PC3 (Fig. 5O,P). However, the level of the exogenous PC3 protein was higher, particularly in the outer EGL (Fig. 5P,V). In parallel, the endogenous PC3 mRNA appeared to be downregulated by the exogenous PC3, suggesting an autoregulatory loop (Fig. 5N,T). A similar pattern was observed in Tg βACT-tTA/TRE-PC3, with the exception that expression of the exogenous PC3 mRNA and protein was stronger and was also very evidently ectopic in the IGL at P1 (Fig. 5C,I,D,J).

In both Tgs, at P1, a strong decrease of BrdU incorporation and of cyclin D1 expression was evident in the outer EGL (in which cyclin D1 is normally expressed) (Shambaugh et al., 1996) (Fig. 5, compare E,F with K,L and Q,R with W,X). This decrease correlated with the expression of exogenous PC3 in the EGL. However, although the decrease in BrdU labeling was evenly distributed, the localization of cyclin D1 was not uniform in all folia and lobules, which raised the possibility that other members of the cyclin D family could be involved. As judged by Western blot analysis of the whole cerebellum, a barely detectable decrease of cyclin D2 expression was observed, whereas cyclin D1 was clearly reduced (10 and 40%, respectively, by densitometry). No change was observed for cyclin D3 and for N-myc, known to exert a positive control on cyclin D2 expression (Knoepfler et al., 2002) (supplemental Fig. 1; available at www.jneurosci.org).

Unexpectedly, an increase of NF160 kDa staining was observed in the whole cerebellum, particularly in the IGL and in the molecular layer, whose thickness was significantly increased (Fig. 6A,F,K,P). Moreover, in some lobules, NF-positive cells were detectable in the inner and even in the outer EGL (Fig. 6A,L, arrows), indicating that GCPs differentiated in an area in which they are normally still proliferating.

To further analyze the differentiation of GCPs at P1, we studied the expression of the basic helix-loop-helix transcription factor Math1, which is normally expressed in GCPs (Akazawa et al., 1995; Ben-Arie et al., 1997, 2000; Helms et al., 2001). In both Tgs, a marked increase in Math1 expression was identified in the outer EGL (Fig. 6B,C and L,M) compared with the control (Fig. 6G,H and Q,R). Math1 was also highly expressed ectopically in the IGL of Tg βACT-tTA/TRE-PC3 and, to a lower extent, of Tg nestin-rtTA/TRE-PC3 (Fig. 6B,C,L,M) (see Fig. 8D,D′), in agreement with the ectopic expression of PC3 protein in this area. Ectopic expression of Math1 protein (and mRNA, data not shown) in the IGL occurred mainly in granule cells but was observed to some extent also in Purkinje neurons (labeled by calbindin) (Fig. 6B,C). Moreover, Purkinje neurons showed abnormal clustering, rather than a coordinated layer, in the Tg βACT-tTA/TRE-PC3 and, to a lower extent, in the Tg nestin-rtTA/TRE-PC3 (Figs. 5D,J,P,V, 6B,G). However, this alteration was not directly correlated with the expression of PC3 in these cells. In fact, PC3 was weakly detectable in Purkinje neurons only in the Tg βACT-tTA/TRE-PC3 (Fig. 5D).

Changes were observed also in the expression of NeuroD, which labels the immediately postmitotic granule cells in the inner EGL and in the molecular layer, thus before and at the onset of their migration (Miyata et al., 1999). These migratory NeuroD-positive granule cells were reduced in number in the molecular and Purkinje layers and appeared to accumulate in the inner and also in the outer EGL in both βACT-tTA/TRE-PC3 and nestin-rtTA/TRE-PC3 Tgs (Fig. 6D,I,N,S), suggesting that granule cell migration was affected. Frequently, NeuroD-positive granule cells in the outer EGL and molecular layer also expressed NF160, indicating an accelerated differentiation (Fig. 6D,N).

Furthermore, an increase was observed in the number of apoptotic nuclei in the outer one-third of the EGL (three to four GCPs layers), as detected by TUNEL assay, in both Tgs at P1, in correlation with the maximal expression of exogenous PC3 protein (Fig. 6E,J,O,T).

We then analyzed cerebella at P5 (Figs. 7, 8). At this stage, in Tg βACT-tTA/TRE-PC3, all of the modifications observed in the P1 cerebellum were still detectable, namely, reduced cerebellar size (Fig. 7A,D), decrease of BrdU incorporation and cyclin D1 expression (Fig. 8 and data not shown), increase of NF160 staining in the IGL (Fig. 7C,F), and increase of Math1 expression in the EGL and IGL (Fig. 7G,J), concomitantly with expression of exogenous PC3 mRNA (Fig. 7B,E). The migration of NeuroD-positive granule cells was still reduced, but the normal localization of Purkinje cells was almost restored (Fig. 7H,K,G,J). In turn, the radial pattern and position of the Bergmann glia, detected by GFAP (Levitt and Rakic, 1980), was severely disorganized at P5 in the βACT-tTA/TRE-PC3 (Fig. 7I,L). Remarkably, however, none of these changes was detectable at P5 in the cerebellum of Tg nestin-rtTA/TRE-PC3, which regained normal morphology (Fig. 7M–T). This correlated with the complete downregulation of exogenous PC3 expression, attributable to the physiological inactivation of the nestin promoter at P5 (Fig. 7N,P).

In summary, in both Tgs, the PC3 protein was overexpressed in the EGL and, ectopically, in the IGL. This was followed by a decrease in EGL proliferation, as was evident by reduced cyclin D1 levels and BrdU incorporation. Moreover, a significant increase in Math1 expression was seen in the EGL, and ectopically in the IGL, with a strong enhancement of granule cells differentiation, as judged by NF and NeuroD levels. The ectopic induction of Math1 correlated with the ectopic expression of PC3 in the IGL and with the disorganization of the Purkinje cell layer. Together, these observations point to an effect of PC3 on the proliferation and differentiation of GCPs as the basis for the cerebellar phenotype.

Cell cycle inhibition, Math1, and NF induction and apoptosis: quantitative analysis in cerebellum

To quantify the changes observed, the Tg mice, βACT-tTA/TRE-PC3 and nestin-rtTA/TRE-PC3, activated since fertilization as described above, were exposed to BrdU for 1.5 hr, and the BrdULI was analyzed in the whole EGL at P1 and P5 (Fig. 8A,A′).

BrdULI was significantly reduced by expression of PC3 in Tg βACT-tTA/TRE-PC3 of 40.5 and 38% at P1 and P5, respectively, whereas BrdULI in Tg nestin-rtTA/TRE-PC3 decreased significantly (37.1%) only at P1, being restored to control values at P5 in correlation with the disappearance of exogenous PC3 expression (Figs. 7N,P, 8A,A′). Measurements of BrdULI were not influenced by apoptotic GCPs, whose nuclei with condensed and fragmented chromatin were not counted (Oberhammer et al., 1992). Cyclin D1 expression was downregulated in parallel with the reduction of BrdULI (Fig. 8B,B′).

The extent of granule cell differentiation was analyzed by measuring the total IGL area positive for NF160 staining. This showed an increase of >2.5-fold in Tg βACT-tTA/TRE-PC3 and in Tg nestin-rtTA/TRE-PC3 at P1, whereas at P5, only the former Tg showed a significant increment (Fig. 8C,C′). A large increase was also observed in the percentage of Math1⁺/total cells in the IGL, up to sevenfold in Tg βACT-tTA/TRE-PC3 and more than three-fold in Tg nestin-rtTA/TRE-PC3 at P1 (Fig. 8D,D′). The increases in NF160 and Math1 expression were highly proportional, suggesting a correlation between them.

As shown in Figure 8, E and E′, TUNEL assay indicated that apoptosis was significantly increased in the EGL of Tg βACT-tTA/TRE-PC3, whereas in the IGL and in Tg nestin-rtTA/TRE-PC3, the increase was slight and nonsignificant. The total number of cells within the EGL (cell number/1 × 10³ μm²), which represents a measure of the GCP pool turnover (see Discussion), did not change significantly in both Tgs with active expression of PC3 (Fig. 8 F, F′). The absence of evident changes in cell number was consistent with the unchanged EGL thickness (Fig. 5 A, G, M, S).

PC3 acts upstream of Math1 and controls its transcription

Having found that transgenic PC3 expression increased the expression of Math1, we sought to test whether Math1 is a target of PC3 action by an independent approach. We therefore generated primary cultures of postmitotic cerebellar granule cells from P8 rats, which do not normally express Math1, grew the cultures for an additional 5 d and transduced them with a recombinant PC3-expressing adenovirus (Adeno-PC3) 24 hr before harvesting. In control cultures infected with Adeno-β-gal, Math1 mRNA was undetectable (Fig. 9A), as expected (Akazawa et al., 1995; Ben-Arie et al., 2000). In contrast, infection with Adeno-PC3 caused a strong reactivation of Math1 mRNA (Fig. 9A). Furthermore, the expression levels of the transcription factors Zic1 (Aruga et al., 1998) and to a lower extent Zipro1 (Yang et al., 1999), which are expressed in vivo in both the EGL and IGL, were also induced by PC3. Similarly, the level of NeuroD mRNA, whose expression is needed for differentiation and is maintained in mature granule cells (Lee, 1996; Miyata et al., 1999), was also slightly increased (Fig. 9A).

Because the level of Math1 increased after infection by Adeno-PC3, we evaluated whether this effect was common to other cell cycle inhibitory genes. We chose the cyclin-dependent kinase inhibitors p21 and p27, known to be expressed in cerebellar granule precursors in the EGL during the period P0–P9 (Shambaugh et al., 2000). No induction of Math1 was detectable after infection of primary cultures of cerebellar granule cells with recombinant p21 or p27 adenoviruses (Adeno-p21 and Adeno-p27), indicating that the induction of Math1 was a specific function of PC3 (supplemental Fig. 2).

Next, we asked whether PC3 could activate Math1 transcription via its promoter elements. For this, we generated the construct pGL3-Math1-pr/-1200, which included the Math1 promoter region upstream to the luciferase reporter gene. This construct was cotransfected with PC3 into primary cultures of cerebellar granule cells or into PC12 cells differentiated into sympathetic neurons by NGF treatment. Indeed, Math1 promoter activity was significantly induced by PC3 transfected in rat P2 (and P8; data not shown) granule cell cultures, as well as in sympathetic neurons (Fig. 9B).

In contrast, when tested in cerebellar granule cells, ectopic PC3 was unable to stimulate a luciferase reporter placed under the control of the Math1 enhancer (construct pGL3-Math1-enh) (Fig. 9B), which is responsible for the positive autoregulation of Math1 (Helms et al., 2000). This suggested that PC3 acts through DNA motif(s) within the Math1 gene promoter region and not through its enhancer (Fig. 9B).

Furthermore, cerebella of E14 Math1^–/– mice, lacking the entire coding region of Math1, showed normal levels of PC3 (Tis21) mRNA, indicating that no feedback mechanism was exerted by Math1 on the transcription of endogenous PC3 and confirming that PC3 acts upstream of Math1 (Fig. 9C).

As a preliminary assessment of the existence of a physiological interaction between Math1 and PC3, we also performed an ISH analysis of the expression of Math1 and PC3 mRNAs in wild-type mice. Their localization throughout the cerebellar development, from E14 to P30, was closely overlapping (supplemental Fig. 3).

The G₁ to S phase transition is inhibited in cerebellar GCPs expressing PC3

In the cerebella of both binary transgenic animals analyzed, overexpression of PC3 correlated with a significant reduction of BrdU incorporation and of cyclin D1 expression, as well as with a marked increase of Math1 and NF160 kDa expression and with apoptosis. The highest expression of exogenous PC3 occurred in the outer EGL, where it was accompanied by massive differentiation of GCPs, and in the IGL, where it was accompanied by ectopic induction of Math1 that is not normally expressed in this area. How can these events be correlated?

Induction of PC3 in the outer EGL of both binary transgenic models inhibited the cell cycle progression of GCPs, as indicated by the reduction of BrdU incorporation, coherently with the observed downregulation of cyclin D1. The decrease of the BrdU labeling index in the EGL could reflect an increase in the fraction of cells in quiescence (Q; i.e., GCPs that became postmitotic), longer duration of the G₁ phase (T_G1) or an increase of the duration of the cycle (T_C). Additional analyses will be necessary to define these possibilities. At any rate, a reduction of cyclin D1 expression, as was observed, invariably leads to an impairment of CDK4 (cyclin-dependent kinase 4) activity and thus to an inhibition of the G₁ to S phase transition (Baldin et al., 1993). Moreover, the primary effect of PC3, as identified by our previous in vitro studies on neuronal and non-neuronal cells, is a selective impairment of G₁ to S phase progression through inhibition of cyclin D1 transcription (Montagnoli et al., 1996; Guardavaccaro et al., 2000; Tirone, 2001). Cyclin D2, which has been shown to play a role in cerebellar development (Ciemerych et al., 2002), displayed only weak reduction, if any, by PC3. This points to cyclin D1, among D-type cyclins, as the main target of PC3 in cerebellum.

Combined influence of PC3 on proliferation, differentiation, and migration of GCPs underlie the cerebellar phenotype

Overexpression of PC3 in the cerebellum at P1 correlated with a widespread increase of the differentiation of GCPs, with fully differentiated granule cells detected even within the EGL, as shown by NF expression. This indicated that the postmitotic state was attained with a higher frequency and that differentiation occurred also before migration.

A strong increase in the generation of postmitotic cells would lead to a decrease of the active pool of proliferating precursors (given that GCPs, after entering Q, migrate outside the EGL within 24–48 hr) (Rakic, 1971), as would a lengthening of the G₁ phase. Indeed, we observed an accumulation of postmitotic NeuroD-positive granule cells in the whole EGL. This may be explained by an excess of differentiating granule cells entering Q and/or by impaired radial migration, as suggested by the disorganization of the Bergmann glia, which guide granule cell migration (Rakic, 1971; Hatten, 1999). Reduced migration, concomitant with the increase of Q and T_G1, might thus account for the observation that the number of EGL cells remained constant.

It is known that the differentiation of granule cells influences the development of Purkinje neurons (Morrison and Mason, 1998), which in turn control Bergmann glia differentiation and the proliferation and migration of EGL granule precursors (Zecevic and Rakic, 1976; Fisher et al., 1993; Sotelo et al., 1994; Dahmane and Ruiz i Altaba, 1999; Kenney and Rowitch, 2000; Komuro et al., 2001). However, it seems unlikely that the disruption observed in the organization of Purkinje cells and of Bergmann glia had a causal effect on the cerebellar phenotype given that, in Tg nestin-rtTA/TRE-PC3 (in which the PC3 protein is undetectable in Purkinje cells at P1), a significant reduction of cerebellar size and granule cell migration occurred in the presence of a mild disorganization of the Purkinje cell layer. This was notably completely reversed at P5, when the overexpression of PC3 in granule cells had ceased. Thus, the altered pattern of Purkinje cells and Bergmann glia (clearly evident only in Tg βACT-tTA/TRE-PC3 in which PC3 is ectopically expressed in Purkinje cells) could have been an additional factor influencing the cerebellar phenotype, partly secondary to the effects of PC3 on granule cells.

Moreover, given that the appearance of the Bergmann glia is not dramatically changed, an alternative interpretation should be considered, which can account for the unchanged EGL thickness without involving granule cell migration. Because Purkinje cells are disorganized and fewer granule cells are generated, the overall volume of the cerebellum is reduced. Thus, if volume and rate of neurogenesis are reasonably matched, then the thickness of the EGL may not be changed, although its area over the surface of the cerebellum is reduced. If, in contrast, we take into account a migration defect in the newly generated granule neurons, it remains to be seen whether this reflects an influence of radial glia or an intrinsic motility problem in the granule cells themselves, whose differentiation has been anticipated.

As a whole, the observations above suggest that the reduced cerebellar size in both bitransgenic models resulted mainly from reduced proliferation of GCPs in the postnatal mice and from the associated increase of the fraction of GCPs entering Q and differentiating. On the other hand, increased apoptosis of GCPs in the outer EGL might have contributed to the reduced cerebellar size, decreasing the number of GCPs. The increased apoptosis observed appears contradictory with previous data, indicating that PC3 protects postmitotic neurons from death caused by deprivation of trophic factors in vitro (Corrente et al., 2002). Notably, apoptosis increased significantly only in the EGL of Tg βACT-tTA/TRE-PC3, in which the expression of PC3 in GCPs was high. We suggest that, in the rapidly dividing GCPs undergoing clonal expansion (Hatten, 1999), a strong inhibition of cell cycle progression represents a conflicting signal that leads to apoptosis, as shown in other systems (Tiemann and Hinds, 1998). The modest increase of apoptosis in the neural tube, or in the IGL of transgenic mice, is consistent with the idea that the excess number of differentiating neuronal stem cells is controlled by programmed cell death, a physiological feature of neuronal development (for review, see Sommer and Rao, 2002).

PC3-dependent induction of Math1 and cell cycle arrest are coordinated but dissociable events

The effects of PC3 on cerebellar development raised the question as to whether GCPs in EGL become postmitotic and differentiate merely as the result of the cell cycle exit triggered by PC3.

In neural crest-derived pheochromocytoma cells, expression of ectopic PC3 led to inhibition of cell cycle in G₁, but this was not accompanied by differentiation. Rather, PC3 strongly enhanced the NGF-induced differentiation of these cells (Corrente et al., 2002; el-Ghissassi et al., 2002; F. Tirone, unpublished data).

In contrast, in cerebellar granule cells, PC3 appears to have an instructive role in GCP differentiation, because we showed that it activates the expression of Math1, both in vivo and in vitro. Although apparently not involved in the process of specification of GCPs, Math1 is essential for their differentiation, as demonstrated by experiments of ablation in vivo and in vitro (Ben-Arie et al., 1997; Gazit et al., 2004). Consistently, its overexpression in vivo has been shown to drive the expression of early differentiation markers in postmitotic granule cells, confirming a role of Math1 in their differentiation (Helms et al., 2001). Thus, the increased differentiation of granule cells associated with the expression of PC3 can be plausibly dependent on the induction of Math1 by PC3. Furthermore, we showed that PC3 can induce the transcription of Math1 by activating the Math1 gene promoter. Such an effect is consistent with the known activity of PC3 as a transcriptional regulator. In fact, PC3, which is in itself devoid of transactivating function and does not directly bind DNA, has been shown to associate with subunits of the multiprotein transcriptional complex CCR4-Not (Rouault et al., 1998; our unpublished data).

The expressions of exogenous and endogenous PC3 mRNA in Tg nestin-rtTA/TRE-PC3 at P1 overlap closely, being maximal in proliferating GCPs of the EGL and rapidly decreasing in deeper regions. In Tg βACT-tTA/TRE-PC3, with a more evident phenotype, exogenous PC3 levels remained high also in mature IGL granule cells. It can thus be assumed that the cerebellar phenotype observed was obtained through various degrees of enhancement of a physiological process, by increased expression of PC3 in granule cell precursors.

We therefore hypothesize that PC3 can physiologically exert a dual action, by promoting the G₁ arrest of GCPs in the EGL through inhibition of cyclin D1 and, in parallel, by stimulating their differentiation through induction of Math1. As a result, GCPs undergo terminal cell cycle exit to quiescence (Fig. 9D). In this model, Math1 may be ineffective, or less effective as a neural inducer, in the absence of a growth arrest signal. Interestingly, it has been observed in Xenopus that the ability of ectopic Xath5 (amphibian ortholog of Math5, a close paralog of Math1) to drive retinal precursors toward retinal ganglion cell fate is enhanced by forcing cell cycle exit via cotransfection of p27 (Ohnuma et al., 2002). On the other hand, in Math1 null mice, which fail to develop the EGL, expression of PC3 in the cerebellar primordium is unaffected, implying that PC3 in itself is not sufficient to induce GCP differentiation. Thus, PC3 may coordinately induce growth arrest and the differentiation activity of Math1 in GCPs. The idea that the induction of Math1 by PC3 may occur physiologically is clearly compatible with the temporal and spatial coincidence of the expression patterns of Math1 and PC3 mRNAs in the developing cerebellum (analyzed at E14 to P30).

Significantly, our gain-of-function transgenics reproduce the cerebellar phenotype of reduced size and foliation seen in the cyclin D1/D2 null and Math1-overexpressing mouse models (Helms et al., 2001; Ciemerych et al., 2002). However, Math1 overexpression in vivo was not associated with decreased BrdU incorporation (Helms et al., 2001), suggesting that its activity requires coordination with cell cycle arrest, which could be exerted by a G₁ checkpoint regulator such as PC3, as we show here (Fig. 9D).

It remains to be verified whether the proposed function of PC3 as a coordinator between cell cycle arrest–exit and differentiation might be more general and applicable to diverse neuronal lineages controlled by different genes that promote neural differentiation, as suggested by the marked prodifferentiative effect of PC3, observed in various CNS and PNS proliferative regions.