Introduction

The mammalian cerebellum is a foliated structure formed of a central vermis and two bilaterally symmetric hemispheres, each with its own sets of fissures [1]. The cerebellum has a regular cytoarchitecture composed by two discrete components: the cerebellar cortex in a superficial position and, buried, an aggregate of neurons that constitutes the deep cerebellar nuclei [2]. The adult cerebellar cortex exhibits three layers. First, the molecular layer located the under the pial membrane comprises two main types of interneurons, the basket and the stellate cells. The Purkinje cell layer contains the cell bodies of Purkinje cells (PCs) whose dendrites arborize in the molecular layer (ML). They are the sole output neuron of the cerebellar cortex, via connections with neurons in the cerebellar nuclei [3], which project efferent fibers to the thalamus, brainstem, and spinal cord [4, 5]. The deepest layer, the granule cell layer, presents the cell bodies of the granule cells (GCs) and other interneurons such as Golgi, Lugaro, candelabrum and unipolar brush cells [6, 7].

Studies using chick–quail chimeras showed that the cerebellum derives from the region of the neural tube, which encompasses the caudal portion of the mesencephalic vesicle and the rostral portion of the metencephalic vesicle [8]. Subsequent gene expression and fate mapping analysis indicated that this region of the CNS arises entirely from the anterior-most rhombomere of the hindbrain, rhombomere 1 [9, 10]. Over this period, it was also seen that the isthmus, the neural tissue at the metencephalic–mesencephalic junction, plays an important role in the establishment of the cerebellar territory [8]. It is accepted that the organizing activity of the isthmic tissue is mediated by the secreted fibroblast growth factor 8 [10, 11].

The establishment of the cerebellar territory is followed by the formation of two germinative compartments, the ventricular neuroepithelium and the anterior rhombic lip. All cerebellar neurons derive from the above regions according to a well-defined spatiotemporal sequence [1]. GABAergic neurons, such as Purkinje cells and inhibitory interneurons, are generated from the ventricular neuroepithelium upon pancreas transcription factor 1a control. By contrast, glutamatergic neurons, including deep cerebellar nuclei neurons, unipolar brush cells and GCs derive from Math-1-expressing progenitors that emigrate from the rhombic lip [12, 13].

There are several mutations that affect the development of the cerebellum (http://www.informatics.jax.org/). One of these is the weaver mutation (Kcnj6 gene), located on mouse chromosome 16 [14, 15]. Its genetic basis consists of a single base-pair substitution that generates a missense mutation (Ser156Gly) affecting the highly conserved H5 domain of the inwardly rectifying GIRK2, a G-protein coupled K+ channel [16].

Homozygous weaver mice (wv/wv) have been proposed as model for Parkinson’s disease [17, 18] and cerebellar ataxia [19] because they present motor abnormalities [20]. These impairments arise as a consequence of the extensive loss of dopaminergic neurons in the midbrain [21, 22] and of GCs, PCs and deep cerebellar nuclei neurons in the cerebellum [23–29]. In the heterozygous weaver mice (+/wv), on the other hand, the depletion of neurons seems to be less evident [23, 24, 26, 27, 30] and they do not present behavioral abnormalities [23, 31].

Previous studies have demonstrated that, in the wv/wv cerebellar cortex, the depletion of GCs and PCs occurs according to a lateral-to-medial gradient of increasing severity [24–27, 29, 32]; the vermis is the most severely affected region, while in the hemispheres the neuronal deficit is spared. This, together with the ectopic location of the surviving GCs and PCs [24, 25, 28, 33], determine important regional differences in the cytoarchitecture of the vermal region [27, 32, 34].

The aim of this paper is to show that the use of several morphological parameters from the cerebellar cortex of control, heterozygous, and homozygous weaver mice in a principal component analysis (PCA)-based cluster analysis [35–38] could be useful to assess cerebellar morphology, as well as to illustrate that these sub-phenotypes exhibit different cytoarchitectonics and foliar pattern.

Materials and Methods

Animals

All mice used in this study were obtained from the colony of control and homozygous weaver mice at Indiana University School of Medicine maintained on the B6CBA-Aw-j/A hybrid stock. The parents of the +/+, +/wv and wv/wv offspring used in the present study were either +/wv or wv/wv females mated to +/wv males. It is worth mentioning that this paper is included in a series which analyzes the development of the +/wv and wv/wv central nervous system. The animal subjects included in the present work were also used in a [3H]TdR autoradiography study to get insight in the generation and settling of +/wv and wv/wv EGL cells and Purkinje neurons [39]. Thus, we have used the series injected at embryonic days E14–15, time at which PC neurogenesis is almost completed, and the label is just residually incorporated. Signs of toxicity were never observed and cerebella of the injected animals appeared normal in both size and cell number.

The mice were maintained in controlled environment (lights on from 07:30 to 19:30 hours; temperature, 22 ± 2 °C). They had free access to food and water. A total of 54 mice were processed for histology at P8. All procedures were approved by the animal care and use committee of our university.

Wv mutation consists of a point mutation in the GIRK2 gene, and was ascertained by means of SSCP (single-strand conformational polymorphism) [40]. Genotypes were also confirmed by microscopic examination of the cerebellum, which is smaller and has disorganized cytoarchitecture as a consequence of the depletion of several neuron populations [24].

Perfusion and Histology

At P8, animals were deeply anesthetized with sodium pentobarbital (50 mg/kg i.p.) and perfused through the heart with 4 % paraformaldehyde in phosphate-buffered saline (pH 7.4). Brains were removed, dissected, and post-fixed for 5 h at 4 °C. Next, they were dehydrated in graded-ethanol solutions prior to being embedded in paraffin. In order to avoid overestimation of cell counts, the block containing the cerebellum was sectioned serially at 10 μm in the sagittal plane and one of every six sections was placed on poly-(l-lysine)-coated slides. Care was taken to keep the long axis of the brain parallel to the side of the block and the bottom of both cerebral hemispheres touching the base of the embedding box, to avoid tilting of the cerebellum so as to obtain true sagittal sections. Only those sections representative of the medial point of the vermis were used in the present study. Sections were stained using two procedures: haematoxylin or cresyl violet.

Morphometry of the Cerebellum

The following features of the cerebellar morphology were quantified per section: length of the cerebellar cortex, area of the EGL, area of the ML, number of EGL cells, and number of PCs. Each parameter was determined in three sections of every experimental animal. Data from each section were added to obtain a mean for each cerebellar feature per mouse. Under identical lighting conditions, images of sectioned cerebella were captured by a CCD-IRIS color video camera (Sony, Japan) coupled to a Zeiss Axiosphot microscope and digitalized. Morphometric analysis was performed with the Visilog 5 software (Noesis, France). Analysis was done as follows: the first step was calibration to convert pixel units to metric units. Subsequently, length of the cerebellar cortex, as well as EGL and ML areas, were delimited manually to obtain initial binary images. Images of total cerebellum, entire EGL and entire ML were submitted to the mentioned software to obtain the length of the cerebellar cortex, area of the EGL and area of the ML.

Counts of EGL cells and PCs were carried out in the pertinent sections by visual scanning of the cerebellar cortex throughout the entire anteroposterior profile. The amount of EGL cells was determined using a 1,000× oil immersion objective and an ocular graticule. Criteria for scoring them included both morphological and staining properties. Small, darkly stained and densely packed cells, most of them round-like in shape were considered as EGL cells. Pyknotic nuclei and endothelial cells were easily discerned. On the other hand, PCs were counted on the bases of several assumptions such as size, the morphological characteristics of the pericarion and distinctive stain properties. These macroneurons were counted at 400×, and considered as present if they possessed large pyriform somata and nucleus with the presence of a distinct nucleolus. Since wv/wv PCs somata are ectopically settled due to disruption of the normal cerebellar cortex cytoarchitecture [27, 32, 34], it may be difficult to distinguish between PCs and any other large neurons in our sections. Total PC counts in the wv/wv were therefore multiplied by a Golgi cell correction factor (85.7 %). This correction factor was derived from the PCs/Golgi cell ratio in lower mammals [41] and it was previously used in quantitative analyses of weaver PCs [26, 42].

Qualitative Analyses of the Cerebellar Cortex

Light microscopic analyses of cerebellum were made with a Zeiss Axiosphot microscope using a wide set of objectives. Observations were focused at the superficial (convex) areas of the following lobules. From the anterior lobe, the lobules lingula, culmen and centralis were examined; the tuber and the pyramis from the central lobe; the uvula from the posterior and the nodulus from the inferior lobe. The deep of the prima, secunda and posterolateralis fissures were also studied. The names of the lobes and lobules used in this paper are those previously assigned by Altman and Bayer [1].

Statistical Analysis

Means ± the standard error of the means were obtained for each quantified parameter of the cerebellar morphology. Pearson correlation coefficient for each pair of variables was computed. To reduce the dimensionality of the variables’ set, a PCA was performed [36]. The new components obtained from the PCA were submitted to a cluster analysis applying the Ward’s Hierarchical Clustering Method [35, 37]. The cubic clustering criterion [43] was used to determine the optimum number of clusters and R 2 distance was used to represent the clusters in a tree dendogram [44].

The accuracy of the classification was evaluated by performing a leave on out cross-validation. The description and comparison of the clusters was analyzed by bivariate analysis of variance; post hoc pair-wise comparisons were carried out using Student–Newman–Keuls method. Additionally, a canonical discriminant analysis [36] was performed to find linear combinations of morphological parameters providing maximal separation among clusters. Total sample standardized canonical coefficients and scatter plots are presented.

The statistical analysis has been performed using software SAS v9.2 (SAS Institute Inc., Cary, NC, USA). P values lower than 0.05 were considered statistically significant.

Results

Assigning Genotypes and Phenotypes

Animals were assigned to one of the three genotypes: wild type (+/+; n = 15), +/wv (n = 28) and wv/wv (n = 11). Assignation was first based on light microscopic criteria in accordance with previous descriptions [24, 25, 45, 46] and then confirmed by SSCP [40].

Quantified parameters of the cerebellar morphology for each genotype are summarized in Table 1. Pearson correlation coefficient for each pair of variables was computed and a correlation matrix was generated (Table 2). Because these variables are highly correlated, a PCA was performed to reduce the number of variables before applying cluster analysis. All variables were standardized such that each variable was given the same weight in the PCA. The first two principal components accounted for 98 % of the total variance (Fig. 1). The first and most important component is an average measure of all variables since all of them show approximately equal loadings. The second component has high positive loadings on the length of the cerebellar cortex and a negative weight on the EGL cell number.

Table 1 Mean ± SEM of the quantified features of the cerebellar morphology for each genotype
Full size table
Table 2 Matrix of Pearson correlation coefficient for each pair of variables
Full size table
Fig. 1
figure 1

Principal components analysis. Position of wild-type (filled circle), heterozygous weaver (empty circle) and homozygous weaver (multiplication symbol) in the plane spanned by the first two principal components. The area delimited by two lines represents the 95 % confidence interval for the two principal components. The first and most important component is an average measure of all variables since all of them show approximately equal loadings. The second component has high positive loadings on the length of the cerebellar cortex and a negative weight on the EGL cell number

Full size image

A cluster analysis was applied to these two principal components in order to classify heterozygous and homozygous weaver mice into potential sub-phenotypes. The cubic clustering criterion indicated that the optimum number of clusters was 5. All controls were classified into cluster 1, heterozygous were categorized into three clusters [2–4], whereas homozygous were initially cataloged into cluster 5 (Fig. 2a). Heterozygous sub-phenotypes were tagged as +/wv 0 (n = 15), +/wv 1 (n = 7) and +/wv 2 (n = 6).

Fig. 2
figure 2

Cluster analysis. Dendogram showing PCA-based cluster analysis results for all of the quantified parameters. Arrows at the side of each tree branch delineate the clusters. a Initial clustering. b Clustering applied to data from homozygous wv/wv animals alone

Full size image

Because many homozygous weaver mice died early, typically before adulthood (personal observation) and much of them are sterile [47, 48], it became difficult to obtain a large group of animals. In consequence, the low number of homozygous animals used in this study (n = 11) could preclude initial classification into clusters and a second analysis was performed considering only this group. Two clusters were then found (5A and 5B, Fig. 2b), designated as wv 0/wv 1 (n = 6) and wv 0/wv 2 (n = 5).

The accuracy in the subsequent leave on out cross-validation of the nested cluster analysis was 100 %. Results from the bivariate analysis of variance and individual data points of each quantified cerebellar feature for all sub-phenotypes are shown in Tables 3, 4, 5, 6, and 7. These analyses revealed that all of the analyzed features vary in accordance to the described sub-phenotypes. These parameters are larger in the +/+, smaller in sub-phenotypes +/wv 0, +/wv 1 and +/wv 2 and smallest in sub-phenotypes wv 0/wv 1 and wv 0/wv 2. When the +/wv sub-phenotypes were separately considered, it was observed that values were greatest in the +/wv 0, less in the +/wv 1, and least in the +/wv 2. Except for the length of the cerebellar cortex, the remainder analyzed parameters were larger in the wv 0/wv 1 than in the wv 0/wv 2 sub-phenotype.

Table 3 Individual data points for the cerebellar cortex length per section in the phenotypes noted in this work
Full size table
Table 4 Individual data points for the external granular layer area per section in the phenotypes noted in this work
Full size table
Table 5 Individual data points for the molecular layer area per section in the phenotypes noted in this work
Full size table
Table 6 Individual data points for the number of external granular layer cells per section in the phenotypes noted in this work
Full size table
Table 7 Individual data points for the Purkinje cell number per section in the phenotypes noted in this work
Full size table

In order to determine which features discriminated the best among clusters, a canonical discriminant analysis was performed. From this analysis, we obtained linear combinations of the variables providing maximal separation among clusters (Fig. 3a and b). The first canonical variable explains more than 95 % of the total variation and the two canonical variables extracted explain the 100 %. Table 8 presents the standardized coefficients for the canonical variables. The number of PCs is the variable which most contributes to the discrimination among the initial clusters, whereas for the further clustering of homozygous animals the number of EGL cells was the one which most contributed to the discrimination between the two clusters found.

Fig. 3
figure 3

Canonical discriminant analysis providing maximal separation among clusters. a Five clusters of mice are well differentiated in the initial clustering. b Two clusters of mice are well differentiated in the isolated clustering of homozygous animals

Full size image
Table 8 Total sample standardized canonical coefficients for the cluster analyses
Full size table

Qualitative Observations in the Cerebellar Cortex

Vermal sagittal sections from each sub-phenotype illustrating the lobular organization of the cerebellar cortex are shown in Fig. 4. It is observed that in each sub-phenotype the cerebellum appears as a folded structure with a pattern of creases and fissures relatively well developed. Despite, compared to the wild-type, foliar size was smaller in the +/wv 0, +/wv 1 and +/wv 2 ones and smallest in the wv 0/wv 1 and wv 0/wv 2 ones. Those fissures that determine the limits between the cerebellar cortex lobes (prima, secunda and posterolateralis) were present in each sub-phenotype at which identification of the cerebellar lobes and lobules was reliable.

Fig. 4
figure 4

Low magnification photomicrographs of the sagittally sectioned cerebellar vermis from wild type (a), heterozygous weaver (b–d) and homozygous weaver (e–f). Photomicrographs b, c, and d correspond to sub-phenotypes +/wv 0, +/wv 1 and +/wv 2, respectively, while those lettered e and f refer to sub-phenotypes wv 0/wv 1 and wv 0/wv 2, respectively. The anterior lobe ranges from the rostral pole of the cerebellar cortex to the fissure prima (long-stemmed arrow), the central lobe between the fissura prima and the fissura secunda (short-stemmed arrow), the posterior lobe between the secunda and the posterolateralis (arrowhead), and the inferior lobe. Vermal lobules are indicated as follows: l lingula, ce centralis, c culmen, d declive, f folium, t tuber, p pyramis, u uvula and n nodulus. Scale bar 1 mm

Full size image

A closer analysis reveals that the +/+ cerebellar cortex exhibits its characteristic trilaminar disposition throughout. Briefly, it presents a subpial structure, the EGL, being formed by small, darkly stained and densely packed cells, most of which are round-like in shape. Immediately beneath the EGL is the ML, which contains basket and stellate cells as well as vertically oriented spindle-shaped cells (the differentiating granule cells). Next, the Purkinje cell layer, composed of large somata of PCs arranged in a single and straight row, is found. Finally, the internal granular layer is formed of very densely-packed somata of small neurons; the GCs.

In contrast, two unusual features are seen in the cerebellar cortex of heterozygous weaver mice. Thus, in the +/wv 0 sub-phenotype PCs are placed in a quasimonolayer in each studied region. This arrangement changed in the sub-phenotypes +/wv 1 and +/wv 2, as the number of strata in which these macroneurons are distributed varied from area to area in both sub-phenotypes (Table 9). Another difference recognized by light microscopy was the presence, in the ML of the +/wv 1 and +/wv 2 sub-phenotypes, of a distinctive band of darkly stained cells, whose density depends on both the sub-phenotype and the studied area (Table 9). Band showing an increased density of cells in the ML was not observed in the +/wv 0 sub-phenotype or in the +/+ one. Figure 5 exhibits both unusual features in the fissure prima and in the uvula of the sub-phenotype +/wv 1.

Table 9 Spatial disposition of Purkinje cells and cellular density in the molecular layer of the +/wv 1 and +/wv 2 phenotypes
Full size table
Fig. 5
figure 5

High magnification photomicrographs of sagittal sections from +/+ (a–b) and +/w 1 (c–d). Photomicrographs a and c correspond to the fissure prima, and b and d to the uvula. egl External germinal layer, ml molecular layer, pcl Purkinje cell layer, igl internal granular layer. Arrows show examples of Purkinje cells. Dashed lines delimit the molecular layer in which a band of darkly stained cells is observed. Scale bar 25 μm

Full size image

When the observations were focused on the weaver homozygotes, it was found that the cerebellar cortex does not exhibit the typical layering in any of the studied areas. The cytoarchitectonic disorder was conspicuous, which is in line with previous reports [27, 32]. This is because the PCs somas are dispersed in several irregular rows, which occupy most of the cortical area with the exception of the EGL and of a small ML. Moreover, the internal granular layer is not present, and GCs are scattered. The EGL, on the other hand, was found throughout the cerebellar cortex. A detailed examination focused in the anterior lobe of the wv 0/wv 2 sub-phenotype revealed that a lot of GCs degenerated within the EGL and completely disappeared. This degeneration was greatest in the culmen, lesser in the centralis and least in the lingula (see Fig. 6 and compare to Table 10).

Fig. 6
figure 6

High magnification photomicrographs of sagittal sections from +/+ (a–c) and wv 0/wv 2 (d–f). Photomicrographs a and d correspond to the lingula, b and e to the centralis and c and f to the culmen. egl External germinal layer, ml molecular layer, pcl Purkinje cell layer, igl internal granular layer. Scale bar 25 μm

Full size image
Table 10 Variation in the EGL cells number per section in three lobules of the anterior lobe
Full size table

Discussion

The results of this study confirm and extend previous data [27, 32, 34]; namely, that there are important regional differences in weaver gene action on the cortical architecture. Moreover, we also show here that these differences determine the existence of several sub-phenotypes in the weaver mutant mouse, showing each one of them impairments both in the cytoarchitectonics of the cerebellar cortical layers and in the foliar pattern. To our knowledge, this is the first time that the aforementioned sub-phenotypes are reported.

A point deserving attention when the sub-phenotypes are considered is whether the differences found both in the parameters quantified as well as the qualitative observations may be due to technical reasons such as mating protocol, exact age of the mice or laboratory procedures. This seems unlikely, as the cerebellar alterations were individually and consistently observed in each sub-phenotype.

In order to obtain even more reliable data connecting the five sub-phenotypes clustered by the presented PCA to regional differences in the cytoarchitecture of their cerebellar cortex, more variables could be considered. Among them, those reflecting differences in the rostrocaudal (ratio of EGL thickness o PCs number among lobes o lobules) or lateromedial axes (rate between the length of the cerebellar cortex or ML area in the vermis vs hemispheres) could be used.

The present qualitative observations show that the ML of +/wv 1 and +/wv 2 sub-phenotypes exhibits a band of darkly stained cells. A similar structure was previously described in the weaver heterozygotes [23, 25]. However, as distinguished from these authors, current findings supply additional information regarding the degree of cellularity in each studied region. Furthermore, our results notice that the presence of this band of cells is independent of: (1) whether a concavity or convexity is considered. For example, it is found both in the deep of the fissure prima and in the top of the pyramis, (2) analyzed lobe. The band is detected both in the anterior and in the posterior lobes, and (3) the neurogenetic timetables of GCs. This is so because the band of cells is observed both in the lingula and in the tuber, two lobules whose GCs have different times of origin [1]. Lastly, we provide evidence that this band is not present in the weaver heterozygotes tagged as +/wv 0 or +/+ sub-phenotypes.

The identity of the cells constituting the band cannot be unequivocally ascertained from our results. However, judging by shape and size of their nuclei, and cytoplasms, we suggest that most of them are young GCs. Moreover, after carefully analyzing sections, it is also proposed that these microneurons are arrested in the ML. This is based on a number of lines of evidence: (1) the rate of GCs migration is slower in the weaver heterozygous than in the +/+ but superior to the wv/wv GCs [23], (2) tissue culture experiments from postnatal cerebella reported impairments in the migratory behavior of +/wv GCs [49, 50], and (3) +/wv ↔ +/+ chimeras demonstrated that the migratory defect in the +/wv GCs is a defect that is to the GCs themselves [33].

Results obtained in the wv 0/wv 2 denote that at the level of the culmen, both the outer proliferative layer and the inner premigratory layer of the EGL exhibit a severe depletion of cells. It is presumed that the same occurs in the remainder studied areas, but, in accordance with our observations, the loss of cells is less intense. As the proliferative layer is disturbed, an important number of GCs are never generated because their precursors die. The decrease in the length of the cerebellar cortex, as well as in the EGL area, observed in the present paper may be related to the loss of GCs precursors. On this matter, it has been documented that the selective elimination of GCs precursors with X-irradiation [1], DNA-modifying agents [51], hyperthyroidism [52] or cytostatic drugs [53] produce a miniaturization of the cerebellum. At the present, it is unknown why GCs precursors are lost. Previous cell ablation experiments [46] and studies of mutant chimeric mice [54, 55] indicate a powerful influence of PCs on the mitotic activity of these neuroblasts.

In addition to the depletion of GCs precursors, we know from the literature that GCs are lost both in the +/wv and wv/wv vermis due to a direct effect of the mutated gene [8, 30, 33, 56]. Two models for GCs death have been proposed [57]. The first claims that continuous depolarization, due to the constitutively active GIRK2wv channel, results in elevated intracellular Ca2+ levels which lead to cells death. The second states that the loss of GIRK2 channel function reduces a major inhibitory pathway in developing neurons, resulting in GC hyperexcitability and causing cell death. Reductions in the ML area observed in the present paper may be related to a lesser density of parallel fibers as a consequence of GCs deficiency.

The deficit of PCs found in this work is in accordance with previous reports [23–27]. How the expression of the weaver gene affects to these macroneurons remains unknown. An attractive hypothesis proposes that one of the aspects of the weaver sub-phenotype is a failure of the cell movements that lead to the fusion of the initially separate bilateral cerebellar anlage, and that this failure to migrate properly leaves some PCs in a position where they are unable to make appropriate connections, leading to their death [27, 28].

In contrast to the wild type, the mutant sub-phenotypes described in the present paper have their foliar patterns altered. Impairments are much more evident in the sub-phenotypes with two doses of the weaver gene (wv 0/wv 1 and wv 0/wv 2) than in those presenting only one (+/wv 0, +/wv 1 and +/wv 2). From these results, the subsequent question is which factors are behind the altered foliation observed when the weaver mutation is present? Nowadays, this is a mystery. In the discussion that follows, we relate GCs, their precursors and PCs as key elements of this enigma. Moreover, in order to propose a model of the weaver gene action on the fissuration pattern, one should go back to the prenatal life to understand what happen at P8.

Earlier research reported that, in normal rodent cerebellum, the smooth cerebellar primordium is divided, at E18.5, by the three principle fissures to generate the four cardinal lobes [1]. Most hypotheses of the cerebellar foliation point to the likely role of mechanical stresses arising from the production of GCs [1, 52, 58]. We have evidences that, around E19, GCs progenitors and PCs loss was detectable in the cerebellar vermis of the weaver homozygotes [59]. The same study also demonstrated that, in mutant embryos, the thickness of EGL and cardinal fissures were smaller in comparison to +/+. Heterozygous weaver affectation was in between +/+ and wv/wv.

As the +/+ cerebellum develops, cardinal fissures grow and lobes are divided into lobules. At P8, most lobules are observed and sublobules have begun to form (Fig. 4a). By contrast, the fissure that determines the limit between the declive and the folium is not present either in those weaver heterozygotes cataloged as +/wv 1 or in the wv 0/wv 1 and wv 0/wv 2 sub-phenotypes (Fig. 4c, e, f). Moreover, in the sub-phenotypes +/wv 0, +/wv 1 and +/wv 2, the centralis is not divided in sublobules (Fig. 4b–d). The same occurs in the wv 0/wv 1 and wv 0/wv 2 sub-phenotypes, which in addition showed the culmen undivided.

From these observations, it is proposed that the reduced foliar pattern observed in our sub-phenotypes might be due to the depletion of GCs precursors and PCs in the prenatal life. We know that in normal mice, the pattern of lobules is regulated after PCs are born [60] and these macroneurons are produced from embryonic days 10 to 14 [39, 42]. When the five sub-phenotypes detected were considered together and were compared to +/+, a consistent pattern emerged. In the sub-phenotype containing more PCs (+/wv 0), the rate of GCs precursor proliferation is higher and more GCs are produced. This sub-phenotype produces a less affected foliar pattern. The +/wv 2 sub-phenotype, on the other hand, presents less PCs and in consequence less GCs are generated because GCs precursor proliferation is lower. The foliar pattern is more affected. Lastly, the wv 0/wv 2 exhibits lesser PCs and, therefore, GCs precursor proliferation is lowest. In consequence the foliar pattern is very affected.

In addition to the possible involvement of GCs proliferation and PCs, it cannot be ruled out that other genes implicated in the cerebellar foliation may be affected by the weaver gene expression [61]. This is because, in our colony, the weaver mutation is maintained by repeated crossing of carrier animals to a C57BL/6 X CBA F1 hybrid mate, and intercrossing of the progeny to generate litters segregating wv/wv homozygotes. These litters will also segregate all other genetic loci at which the C57BL/6 and CBA strains carry different alleles. This uncontrolled and undefined genetic variability may be a factor involved in the sub-phenotypic differences observed. In the future, experiments can be designed to identify gene expression differences that correlate with the sub-phenotypes reported here.

Further work on the weaver mutant may shed new light on how this mutation impairs the cytoarchitecture of the cerebellar cortex, and may be an excellent animal model for studying mechanisms involved in the cerebellar foliation.

Conclusion

The coexistence of several sub-phenotypes in the weaver mutant is reported for the first time and they have been reliably categorized by PCA-based cluster analysis of several morphological parameters from the cerebellar cortex.