Maintenance of Wnt-3 expression in Purkinje cells of the mouse cerebellum depends on interactions with granule cells

In the mammalian nervous system, the fate of neuroepithelial cells is determined by cell-cell interactions and by intrinsic programs (McConnell and Kaznowski, 1991; Watanabe and Raff, 1990). The factors that affect cell fates are largely unknown, due in part to the complexity of the nervous system and to difficulties in manipulating neuronal cells.

The cerebellum is one of the more simple areas of central nervous system, having three defined layers and a limited number of well-described cell types in the cortex (Ito, 1984). In addition, the mouse cerebellum develops extensively during postnatal life, facilitating manipulations and transplantation experiments (Snyder et al., 1992). Moreover, numerous mouse mutants that affect cerebellar development and function are available (Lyon and Searle, 1989).

The cerebellar cortex contains five types of neurons: Purkinje, granule, basket, stellate and Golgi cells. The Purkinje cells, being the sole output of the cerebellar cortex, form a monolayer between the molecular layer and the internal granular layer (Altman, 1972b). Purkinje cells are generated during embryonic life but their maturation, formation of the dendritic tree and synaptogenesis, occurs during postnatal life (Altman, 1972b). During normal development, granule cells proliferate in the most external layer of the cerebellar cortex, the external granular layer (EGL). At postnatal stages, the granule cells exit the cell cycle and migrate inward to the deeper layer of the cerebellar cortex to form the mature granular layer, just below the Purkinje cell layer. Their axons are left behind in the molecular layer as parallel fibers. After arrival in the granular layer, granule cells start to make synaptic contacts with their postsynaptic targets, the Purkinje cells.

Pharmacological and irradiation ablation studies have determined the influences of particular cell types in the formation of the mature cerebellar cytoarchitecture. In addition, studies on cerebellar mutant mice have been particularly informative with respect to the role of cell-cell interactions during neuronal development. By these approaches, it has been established that climbing fibers and parallel fibers make synaptic contacts with Purkinje cells and regulate Purkinje cell maturation (Ito, 1984), but the molecular factors involved in these developmental interactions are unknown.

One approach to identify the molecules involved in these processes is to analyze the expression of genes encoding developmental signaling factors. The Wnt gene family, encoding cysteine-rich secreted proteins, is involved in the development of the Drosophila and mouse embryos (review in McMahon, 1992; Nusse and Varmus, 1992). In Drosophila, wingless (wg), a member of the Wnt gene family, mediates the interactions between wg-expressing cells and the neighboring cells expressing the homeobox gene engrailed (en). The expression of wg and en is essential for the proper patterning of the embryo such that mutations in either gene result in change in cell fate and cell death (Ingham and Martinez Arias, 1992). It has recently been shown that wg is also required for proper specification of neuroblasts in the Drosophila CNS (Chu-Lagraff and Doe, 1993). In the mouse, Wnt-1 null mutations result in the loss of the midbrain region and the cerebellum (McMahon and Bradley, 1990; Thomas and Capecchi, 1990). One of the first manifestations of the Wnt-1 phenotype is a decay in the mouse engrailed gene (En) expression in the most anterior region of the midbrain, suggesting a role for Wnt-1 in the maintenance of En expression (McMahon et al., 1992).

At least ten members of the Wnt gene family have been isolated in the mouse (Gavin et al., 1990; Roelink and Nusse, 1991). Many of these genes are expressed during embryogenesis in specific areas of the CNS (Gavin et al., 1990; Roelink and Nusse, 1991; Salinas and Nusse, 1992; Parr et al., 1993). For example, Wnt-3 is expressed in a restricted area of the diencephalon, in the D2 neuromer. Expression of this gene appears before morphological signs of segmentation, suggesting that Wnt-3 provides positional information for the specification of the neuromers (Salinas and Nusse, 1992; Parr et al., 1993). In addition, Wnt-3 is expressed in the developing cerebellum, from embryonic to fetal stages, suggesting a possible role in early development of this structure (Salinas and Nusse, 1992). Based on that observation, we have examined the pattern of expression of Wnt-3 during cerebellar development, seeking to find a model system to study the role of Wnt genes in cell-cell interactions during CNS development. We show that Wnt-3 is expressed in the Purkinje cell layer and that its expression is modulated by interactions with presynaptic neurons, the granule cells.

Embryos were obtained from mating of CD1 mice. The day of plug formation was designated as embryonic day 0.5. Adult tissues were obtained from CD1 mice and from weaver or wild-type animals generated by mating of B6CBACa-A^w-J/A-wv mice. meander tail mutant animals were obtained by mating of C57BL/KsJ mea^2Jm +/+ animals. The genotype of the animals were established by their behavior and the size of the cerebellum. Adult animals were perfused with 4% paraformaldehyde (PFA) in PBS and their brains further fixed overnight in the same buffer at 4°C. Embryos were dissected from membranes and fixed in 4% PFA overnight at 4°C. The tissues were washed in PBS, saline solution and dehydrated through ethanol/saline solutions and stored in 100% ethanol. After treatment with xylene, the specimens were embedded in paraffin and sections of 8 μm were generated.

In situ hybridizations were performed according to Wilkinson (Wilkinson et al., 1987) with some modifications (for detail see Salinas and Nusse, 1992). The Wnt-3 probe was a 1.2 kb SalI-EcoRI fragment corresponding to the non-coding and non-conserved region of exon 5 of the gene (Roelink and Nusse, 1991). En-2 probes were generated according to Davis et al., 1988. The calbindin probe was an 800 bp rat cDNA (Lomri et al., 1989). Using partial alkaline hydrolysis 500 nucleotides probes were generated (Cox et al., 1984). Specimens were processed according to Salinas and Nusse (1992).

Detection of Calbindin protein was performed using a monoclonal anti-Calbindin-D from Sigma Chemical Co. Paraffin sections were treated in xylene and rehydrated through ethanol solutions up to PBS. After preabsorption with 5% normal horse serum, the specimens were incubated for 2 hours with the anti-Calbindin monoclonal antibody diluted 1:200. The specimens were extensively washed in 0.2% Tween PBS, incubated with 5% horse serum and then with anti-mouse biotinylated antibody diluted 1:200 in 5% normal horse serum. After several washes in 0.2% Tween PBS, the specimens were treated with (1/6) 5% H₂O₂ in methanol for 15 minutes and washed in PBS. The detection reaction was performed according to Vectastain ABC protocol. Specimens were dehydrated through an increasing series of ethanol solutions, treated with xylene and mounted.

Using in situ hybridization techniques, we found that Wnt-3 is expressed in the Purkinje cell layer of the adult cerebellum (Fig. 1A), pontine nuclei (Fig. 1B), inferior olivary nuclei (Fig. 1C), lateral nucleus, medulla oblongata and hypoglossal nuclei (Fig. 1D), and lateral cochlear nuclei (Fig. 1E). It is of interest that Wnt-3 is expressed in several of the nuclei that form part of the basic cerebellar circuitry, such as the ponti and reticular nuclei, the source of afferent mossy fibers; and the olivary nuclei, the source of climbing fibers. Wnt-3 is also expressed in several nuclei of the dorsal thalamus as well as in the hippocampus (Salinas and Nusse, 1992). The expression pattern of Wnt-3 in the adult brain is summarized in Table 1.

The expression of Wnt-3 in the Purkinje cell layer of the cerebellum prompted us to examine the pattern of Wnt-3 expression during the development of the cerebellum. Our previous analysis showed that Wnt-3 appears at embryonic day 8.5 in the midbrain-hindbrain region of the developing mouse embryo (Salinas and Nusse, 1992), an area that gives rise to the cerebellum. At embryonic day 11.5, Wnt-3 is expressed in the developing cerebellum (Fig. 2A), while at embryonic day 13.5, the gene is expressed in cells located close to the ventricular zone (Fig. 2B, arrow). During fetal development, Wnt-3 expression was detected in a diffuse area of the developing cerebellum (Fig. 2C). At that stage, Wnt-3 expression was also observed in the developing inferior olive, dorsal nuclei (Fig. 2C) and in the developing deep cerebellar nuclei (Fig. 2C).

Because the development and maturation of Purkinje cells proceed during postnatal life, we examined the pattern of expression of Wnt-3 in the postnatal cerebellum. Low levels of Wnt-3 expression were detected in a group of cells located close to the pial surface of the cerebellum of the newborn mouse (Fig. 2D), the area where the Purkinje cell layer starts to form. At postnatal day 6, Wnt-3 is clearly expressed in the Purkinje cells (Fig. 2E) in a regionally different pattern with high levels in the posterior region and low levels in the anterior region of the cerebellum (Fig. 2E). Wnt-3 expression increases until P15 when adult levels are reached (Fig. 2F). The dendritic tree of the Purkinje cells develops from postnatal day 5 and reaches maturity at postnatal day 21 (Altman, 1972b). Therefore, the increased expression of Wnt-3 at postnatal stages coincides with the maturation of the Purkinje cells.

Wnt-3 expression in not restricted to the Purkinje cell layer, as expression is also seen in the deep cerebellar nuclei of newborn and adult animals (Fig. 2D,E) and at low levels in the white matter (P15 and adults; Figs 4B, 2F).

Parallel and climbing fibers have profound effects on the arborization of the Purkinje cell dendritic tree (Ito, 1984). Climbing fibers start to make contacts with the somata of Purkinje cells at birth (Mason et al., 1990) and parallel fibers make transient contacts with Purkinje cells by postnatal day 5 (Altman, 1972b). Because Wnt-3 expression increases from P6 to P15 in Purkinje cells, we decided to test the possible influence of the granule cells on Wnt-3 expression in Purkinje cells by examining the agranular cerebellum of the weaver mutant mice (Goldowitz, 1989). In this mutant, granule cells proliferate in the external granular layer (EGL), but then fail to migrate to the internal granular layer (IGL) and subsequently degenerate (Hatten et al., 1986; Rakic and Sidman, 1973). To follow the fate of granule cells, we used En-2 expression as a marker (Davis et al., 1988). En-2, a homeobox gene, is expressed in the developing cerebellum and in the adult granular layer of the cerebellum (Davis et al., 1988).

En-2 expression was detected in the EGL and in the growing IGL of P15 wild-type animals (Fig. 3A). In weaver mutant cerebellum, En-2 expression is limited to the external germinal layer, not unexpected, because of the failure of weaver granule cells to migrate to form the IGL (Fig. 3C) (Rakic and Sidman, 1973). Wnt-3 is expressed in the Purkinje cells of both the wildtype and weaver cerebellum (Fig. 3B,D).

At later stages of development, expression of En-2 was only detected in the IGL of the wild-type cerebellum (Fig. 3E). At this stage, P22, all granule cells had migrated into the IGL and only vestigial granule cells can be observed in the pial membrane of some animals (Altman, 1972a). In the weaver cerebellum, En-2 expression disappears from the external layer, concomitant with the disappearance of the EGL (Fig. 3G). Although most weaver granule cells die in the EGL before migration to the granular layer, there are areas in the cerebellum where some granule cells migrate and form the granular layer. Interestingly, we observed low levels of En-2 expression in the flocculonodular lobe (Fig. 3G, arrow). Wnt-3 is expressed in the Purkinje cell layer of the P22 wild-type cerebellum (Fig. 3F) while weaver Purkinje cells express Wnt-3 at much lower levels (Fig. 3H).

At P28, expression of En-2 was observed in the mature granular layer of the wild-type cerebellum (data not shown) but no expression of En-2 was detected in homozygous weaver cerebellum except in the flocculonodular lobe (Fig. 3I, arrow). Wnt-3 expression is almost absent in the Purkinje cells of weaver animals, although some expression was observed in the flocculonodular lobe (Fig. 3J, arrow).

These experiments show that Wnt-3 expression is not maintained in the Purkinje cells of the weaver mutant and that the decline is concomitant with the disappearance of granule cells. The finding that Wnt-3 expression is maintained where vestigial granule cells are present, such as in the flocculonodular lobe, suggests that the maintenance of Wnt-3 expression depends on the presence of granule cells.

To test whether the disappearance of Wnt-3 expression in Purkinje cells of weaver mutant animals was due to cell degeneration, we examined the expression pattern of Calbindin-D protein (Iacopino et al., 1990; Wuenschell and Tobin, 1988), a Purkinje cell marker present in the somata and dendrites (Mason et al., 1990; Wassef et al., 1985). Calbindin-D expression was examined in sections adjacent to those used for the analysis of Wnt-3 and En-2 expression. At P15, Calbindin-D expression was observed in the Purkinje cell body and dendrites of wild-type and weaver animals (Fig. 4A,B). At P22, Calbindin expression was detected in the almost mature Purkinje cells of the wild-type cerebellum (Fig. 4C). In the weaver cerebellum, Calbindin-D is also expressed in Purkinje cell body and dendrites (Fig. 4D), although a disorganized Purkinje cell layer is evident and the dendritic tree is less developed (compare Fig. 4C,D). At P28, expression of Calbindin-D reveals the mature dendritic tree of the wild-type Purkinje cells (Fig. 4E). In the weaver cerebellum, Calbindin-D is expressed in the cell body and dendrites, although a poorly developed dendritic tree is evident (compare Fig. 4E,F). Thus, Purkinje cells from weaver mutant cerebellum are not degenerated at P22 and P28, showing that the decreased expression of Wnt-3 is not due to cell death.

Several studies have shown that the granule cells are the target of the weaver mutation (Gao et al., 1992; Hatten et al., 1986). The effect on Purkinje cell maturation, manifested by a poor dendritic tree, is considered to be a secondary effect due to the lack of granule cells (Goldowitz, 1989). Accordingly, the decline of Wnt-3 expression in Purkinje cells is most likely due to the lack of interactions with granule cells.

The meander tail mutation also lacks granule cells in the mature cortex (Ross et al., 1990). The remarkable feature of this mutation is that the lack of granule cells and the disorganized Purkinje cell layer are restricted to the anterior lobes while the posterior lobes are completely normal (Ross et al., 1990), prompting us to examine Wnt-3 expression in meander tail.

Expression of En-2 shows that the anterior lobes lack granule cells while the posterior lobes express En-2 in the mature granule cell layer (Fig. 5A). Analysis of Wnt-3 expression in P33 meander tail animals shows that Wnt-3 is expressed in the Purkinje cell layer of the posterior lobe of the cerebellum (Fig. 5B) but only a few Purkinje cells are positive in the anterior lobe (Fig. 5B). The lack of Wnt-3 expression in the Purkinje cells of the anterior lobe is not due to cell degeneration since high levels of Calbindin expression were found in these cells (Fig. 5C). Interestingly, Purkinje cells at the anterior and posterior boundary that come to lie close to the granule cells of the posterior region do express Wnt-3 (Fig. 5A, arrow). The decay of Wnt-3 expression in the weaver and the meander tail mutant animals was confined to the Purkinje cells layer and was not seen outside the cerebellum, supporting the notion that this decline is specifically caused by the lack of granule cells.

In Drosophila, there is good evidence for the interaction between wg, a member of the Wnt gene family and the homeobox gene en. This interaction occurs between cells that are adjacent to each other. en is initially regulated by the pairrule genes, but maintenance of its expression is dependent on the expression of wg in adjacent cells. In turn, maintenance of wg expression requires correct en expression (DiNardo et al., 1988; Heemskerk et al., 1991; Martinez Arias et al., 1988) and these mutual interactions are essential for correct patterning of the embryo. The similarities between this configuration and the cerebellum where two interactive cell populations, Purkinje and granule cells, express Wnt-3 and En-2 respectively, raises the possibility that the effect of granule cells on Wnt-3 expression in Purkinje cells would be mediated by En-2. To test this, we examined Wnt-3 expression in En-2 mutant animals generated by gene targeting (Joyner et al., 1991).

However, Wnt-3 expression was the same in heterozygous animals and in En-2/En-2 homozygous mutant animals (Fig. 6A,B), although the Purkinje cell layer seems to be irregular in the homozygous mutants (Fig. 6B). Therefore, En-2 alone does not mediate the effect of granule cells on the maintenance of Wnt-3 expression.

We have shown that a member of Wnt growth factor family, Wnt-3, is expressed in a defined cell population of the cerebellum, the Purkinje cells and that maintenance of its expression depends on the presence of their presynaptic neurons, the granule cells. Here we discuss the possible role of Wnt-3 during neuronal maturation.

During early stages of development, E 8.5, Wnt-3 expression is restricted in the midbrain-hindbrain region, suggesting that Wnt-3 may play a role in early patterning of this area (Salinas and Nusse, 1992). At E13.5, Wnt-3 expression becomes restricted to a subset of cells in the developing cerebellum, located adjacent to the ventricular zone. Tritiated thymidine tracing experiments have shown that Purkinje cells are located in this area after their last mitotic division (Altman and Bayer, 1985). At fetal stages Wnt-3 expression becomes diffuse, although some expression is evident at the cortical areas. At birth, low expression is observed in a strip of cells in the cortex that coincides exactly with the newly forming Purkinje cell layer. Although we cannot establish whether the gene is actually expressed in the Purkinje cells at these stages, Wnt-3 expression follows the trajectory made by Purkinje cells from the time they are born to the time they migrate and settle into the cortical areas of the cerebellum.

Wnt-3 expression starts to increase in the Purkinje cell layer from P6 until P15, when it reaches its highest level. This period coincides with phase II of Purkinje cell maturation as defined by Altman, when Purkinje cells form transient apical dendritic cones and perisomatic processes (Altman, 1972b). At P7, the Purkinje cells start to develop primary dendrites in the nodulus, the most posterior part of the cerebellar vermis, whereas only a few dendrites are observed in more anterior regions (Altman, 1972b). Interestingly, we observed that Wnt-3 expression is higher in the nodulus and the uvula than in the anterior region of the cerebellum. The increase in Wnt-3 expression coincides with the formation of synaptic contacts between climbing fibers and Purkinje cells. Around the same time, the first contacts are made between granule cells and Purkinje cells (although they do not make synapses yet; Altman, 1972b).

En-2 is expressed in the EGL and IGL of the cerebellum providing a good molecular marker for granule cells. In weaver animals, En-2 is only expressed in the EGL, although low expression was detected in the flocculonodular lobe, suggesting that some granule cells in that area are able to migrate. The flocculonodular lobe corresponds to the oldest phylogenetic area of the cerebellum named the archicerebellum (Ito, 1984). Although regional differences in the capacity of weaver granule cells to migrate into the IGL have been described (Herrup and Trenkner, 1987), the flocculonodular lobe was not previously detected as one of these regions. The lack of good granule cell markers may have impaired the detection of these regional differences. The expression of En-2 in weaver granule cells in the flocculonodular lobe suggests that these cells may be different from the rest of the granular layer. This regional diversity may arise from differences in the time of origin of these cells or from differences in the environment of the flocculonodulus, resulting in local migration of weaver granule cells.

Although we have used En-2 as a granule cell marker for in situ hybridizations, it is important to note that En-2 is also expressed in some cells of the molecular layer as detected in transgenic animals carrying the En-2 promoter fused to lac-Z (Logan et al., 1993) and using specific antibodies against the En protein (A. Joyner, personal communication). The expression in the molecular layer is not detected by in situ hybridization, possibly due to the low sensitivity of this technique as compared to the detection of Lac-Z expression.

The weaver mutation is a non-cell autonomous defect of the granule cells (Gao et al., 1992). Analysis of agranular cerebellum of the weaver mutant shows that Wnt-3 expression is present at P15, but its expression is not maintained and disappears by P22. Analysis of Calbindin-D protein expression indicates that cell degeneration is not apparent in the Purkinje cell layer of weaver animals at this stage. The decline of Wnt-3 expression occurs at the time when the Purkinje cell dendritic trees are more severely affected, that is between P22 and P28. One interpretation of these results is that the maintenance of Wnt-3 expression depends on the presence of normal granule cells. Another hypothesis is that the weaver mutation directly affects Purkinje cells, which is manifested by the decline of Wnt-3 expression. We consider this unlikely, because chimeric experiments have shown that Purkinje cells derived from weaver mutant animals develop normally when surrounded by wild-type granule cells (Goldowitz, 1989). In addition, we found areas of the weaver cerebella where Purkinje cells do express Wnt-3, in the flocculonodular lobe. This region has granule cells that have migrated from the EGL, as determined by the expression of En-2. The meander tail mutant, characterized by the absence of granule cells and a disorganized Purkinje cell layer of the anterior cerebellum, provides further evidence that Wnt-3 expression in Purkinje cells depends on the presence of granule cells.

What are the factors produced by granule cells that regulate Wnt-3 expression? Little is known about regulation of Wnt gene expression in vertebrates (McMahon et al., 1992). More is understood about the Drosophila embryo, in which the maintenance of wg expression is regulated by en in the epidermis. This regulation is accomplished by a loop of interactions between the wg- and en-expressing cells, located just adjacent to each other (DiNardo et al., 1988; Heemskerk et al., 1991; Martinez Arias et al., 1988). A somewhat similar pattern is seen in the cerebellum where Wnt-3 and En-2 are expressed in adjacent cells which interact with each other. Despite this similarity, our analysis of the En-2/En-2 mutant generated by homologous recombination (Joyner et al., 1991) shows that En-2 does not mediate the effect of granule cells on Wnt-3 expression. There are gaps in the expression of Wnt-3 in some areas, but these are due to irregularities in the Purkinje cell layer (not shown).

What is the function of Wnt-3 in Purkinje cells? In the weaver and meander tail mutants, Wnt-3 expression is not maintained in those Purkinje cells that are not in contact with granule cells. The same Purkinje cells fail to undergo the final steps of differentiation, as shown by the poorly developed dendritic tree and a disorganized mature monolayer. Our results suggest that Wnt-3 plays a role in the dendritic differentiation of the Purkinje cells and that Wnt genes in general are not only involved in early patterning of the CNS but that they also regulate neuronal maturation.

The authors thank Drs Carol Mason, Sue McConnell and members of our laboratories for comments on the manuscript. We are grateful to Dr Alexander Joyner and Kathleen Millen for providing the En2/+ and En-2/En-2 material and the En-2 probe. This work was supported in part by the Leukemia Society of America (P. C. S.), by the Howard Hughes Medical Institute and by the National Cancer Institute, DHHS under Contract N01-C0-74101 with ABL; R. N. is an investigator of the Howard Hughes Medical Institute.