Animals analysed in the present study will be referred to as postnatal when a typical EGL is present in the cerebellum (from birth to 5 weeks of life), peripuberal when a SPL is detectable on the cerebellar surface (from 2 to 5 months), and adult (1–3 years old rabbits, devoid of cerebellar SPL). Although in rabbits puberty occurs around the fourth month of life, animals aged 2–5 months were considered to be part of a homogeneous group due to the occurrence of the SPL. Most qualitative and quantitative analyses performed on peripuberal rabbits are referred to the three-month old subjects, and differences are reported concerning older animals at specific stages.
Newly-formed cells with specific morphologies are largely present in the cerebellar cortex of peripuberal rabbits
We have previously shown that the cerebellum of peripuberal rabbits is characterized by a proliferative superficial layer (SPL) which contains chains of newly generated neuroblasts () 
. In addition, by using markers of cell proliferation the occurrence of many newly generated cells is also visible within the cortical layers, a substantial amount of these cells being still detectable two weeks after their birth 
. In order to visualize the morphology of these cells, we employed antibodies raised against markers of structural plasticity, namely the polysialylated form of the neural cell adhesion molecule (PSA-NCAM) and the collapsin response mediator protein-4 (CRMP-4) (). In spite of the progressive reduction of the SPL during the fourth and fifth month of the rabbit postnatal life, followed by its disappearance at the sixth month, many cells expressing the above-mentioned markers were still detectable in the cortex, particularly in the molecular layer (; ). Staining with collapsin response mediator protein-4 (CRMP-4), although fainter than polysialylated form of the neural cell adhesion molecule (PSA-NCAM), revealed cells falling into three main morphological types: bipolar, polarized neuronal-like and multipolar (). The polysialylated form of the neural cell adhesion molecule (PSA-NCAM) was detectable in bipolar and polarized, but not in multipolar cells. On the other hand, multipolar cells showed a stellate, highly ramified morphology and were immunoreactive for the cytoskeletal protein Map5 (Map1B; ) 
Newly generated cells in the cerebellar cortex of peripuberal rabbits.
Number of cells (cells/10 mm) immunoreactive for PSA-NCAM and Map5 in the peripuberal and adult rabbit cerebellar cortex.
Quantitative analyses showed that the bipolar/polarized PSA-NCAM+ cell population represents about 90% (580 cell/10 mm +/− 84) on 10% of multipolar, Map5+ cells (60 cell/10 mm +/− 25), with respect to all cortical cells identified by these markers (see ).
Bipolar cells were mainly radially-oriented, showing the typical morphology of migrating elements, with a leading and a trailing process (). Polarized neuronal cells were observed in the molecular layer with no particular orientation (). They showed a piriform cell body (15–20 µm long), a long, thin axonal-like process with a U-shaped direction remaining within the molecular layer, and one to three thicker, dendritic-like processes on the opposite side of the cell (). Quantitative analysis showed that most PSA-NCAM+ cells (61%; 360+/−38,9 cells/10 mm see and ) reside in the molecular layer, the remaining percentage being widespread in the granule cell layer (35%; 200+/−42,1 cell/10 mm) and, to a lesser extent, in the white matter (4%; 20+/−2,9 cells/10 mm) (). In addition to the typical bipolar and neuronal-like morphologies, many intermediate forms characterized by an axonal-like process on one side and some dendritic branches on the opposite side () were frequently observed (the relative percentages with respect to the total are given in ). The complete set of morphologies, including the neuronal shape, was detectable exclusively within the molecular layer ().
The Map5+ multipolar cells were characterized by a spheric cell body, about 8–10 µm in diameter. These cells showed many, ramified processes reminiscent of the stellate-shaped elements known as synantocytes 
. Those located in the proximity of Purkinje neurons showed a flattened shape when cut transversely (). In contrast with PSA-NCAM+ cells, they were more homogeneously distributed in both the molecular and granule cell layer (see and ; ML: 29,9 +/− 11,9 cell/10 mm; GL 27,5 +/− 11,9 cell/10 mm) and, to a lesser extent, in the white matter (2,3 +/− 1,4 cell/10 mm).
Morphological observations and countings carried out on both sagittal and coronal sections revealed that PSA-NCAM+ and Map5+ cell populations were homogeneously distributed in all dorsal/ventral and medial/lateral lamellae. To test which among these cells are newly generated we performed double labellings with systemically-administered 5-bromo-2'-deoxyuridine (BrdU) then revealed at different post-injection survival times. As previously described 
, after 2 hours (h) most cell proliferation was observed to occur in the SPL, whereas at subsequent survival times (5–15 days) double stainings of 5-bromo-2'-deoxyuridine (BrdU) with PSA-NCAM and Map5 respectively, clearly revealed that a large population of cells expressing such markers in the peripuberal rabbit cerebellar cortex were newly generated ( and Figure S1A
Newlyborn elements in the cerebellar cortex of peripuberal rabbits belong to distinct populations of neuronal and glial cell precursors
Neuroepithelial-derived interneuronal precursors
The group of PSA-NCAM+ cells including bipolar, intermediate and polarized neuronal morphologies was also immunoreactive for the microtubule binding protein doublecortin (DCX) 
, thus confirming they actually represent a neuronal population (). In the rabbit cerebellar cortex, PSA-NCAM and doublecortin (DCX) actually identify an almost overlapping cell population (). Accordingly, in the BrdU/PSA-NCAM double stainings carried out at 5, 10, and 15 days post-injection survival times () the occurrence of bipolar double-labelled cells was more frequent at early survival times, whereas that of more differentiated morphologies (neuronal-like) was detectable at subsequent survival times (see Figure S2A
). This indicates that a neurogenic sequence leading to a young neuronal cell type in the molecular layer does occur during a two week temporal window ().
Newly generated, neuronal-shaped cells in the cerebellar cortex of peripuberal rabbits.
In order to further explore the nature and origin of this population of newly generated cells we combined the immunolocalization of the above described markers with different transcription factors involved in the specification of cell progenitors. The transcription factor Pax2, namely a marker for GABAergic cerebellar interneurons of neuroepithelial origin [8;17], was detectable in PSA-NCAM+/DCX+ cells belonging to all morphological types (). Pax2 staining was generally weaker in ramified cells compared with bipolar ones, thus being probably down-regulated as differentiation proceeds, as confirmed by the incidence of BrdU+/Pax2+ double stainings (see Figure S2B
). The same cell population was consistently negative for both the transcription factors Sox2 and Olig2 (). In addition, most polarized neuronal-like cells showed a remarkable staining for the neurotransmitter γ-aminobutyrric acid (GABA, ), another marker for cerebellar cortex inhibitory interneurons which is concentrated mainly in young neurons before they become terminally differentiated 
. No γ-aminobutyrric acid (GABA) immunoreactivity was detectable in bipolar-shaped cells, nor in the SPL.
By gathering data of co-expression of cell proliferation markers (at 2 h–15 days survival) with markers for structural plasticity PSA-NCAM, collapsin response mediator protein-4 (CRMP-4), and DCX, the transcription factor Pax2, and the neurotransmitter γ-aminobutyrric acid (GABA) in newlyborn cerebellar cells (summarized in ), it can be concluded that a subset of GABAergic interneurons are generated within the molecular layer of the peripuberal rabbit cerebellum.
None of the newly generated cells were immunoreactive for specific markers of common cerebellar neuronal cell types, such as calbindin, calretinin, or parvalbumin (), during the first two weeks after their birth.
Distribution of other* antigens in newlyborn and non-newlyborn cell populations in the cerebellar cortex of peripuberal and adult rabbits.
Multipolar, oligodendrocytic-like cell precursors (synantocytes)
The multipolar, Map5+ cells () were immunoreactive for CRMP-4 but not PSA-NCAM, DCX, or GFAP ( and ; ). They were consistently double-stained for both the transcription factors Sox2 and Olig2 [19;20] () but not for Pax2 (). Thus, the distribution of a combination of cell markers and transcription factors suggests that the PSA-NCAM+/DCX+ cells and the Map5+ cells actually represent two distinct populations, the former being far more numerous than the latter, with a 9/1 ratio in the three-month old animals. Student's Test-T revealed that the number of PSA-NCAM+ cells in 3 month-old rabbits is significantly higher than the number of Map5+ cells (ML p
0,003 GL p
0,019; and ).
Newly generated, multipolar cells in the cerebellar cortex of peripuberal rabbits.
Using BrdU/Map5 double stainings () many Map5+ multipolar cells (Figure S1A
) were found to be generated, starting from the second week survival time after the injection of the DNA marker. No particular spatio-temporal pattern was detectable in these newlyborn cells by considering different morphologies at different survival times and in different cerebellar layers (). Unlike PSA-NCAM+/DCX+ cells, none of the Map5+ cells were found to be double-stained for BrdU at early (1 to 5 days) post-injection survival times, thus indicating that Map5 starts to be expressed later (see Figure S1A,B
The immunocytochemical characterization of multipolar cells with transcription factors, markers of cell proliferation and neurogenesis, indicated that in the peripuberal rabbit cerebellar cortex, in addition to molecular layer interneurons another population of newly generated cells involves glial-like cell progenitors. These cells are reminiscent of synantocytes 
, a recently described cell type sharing a morphology of protoplasmic astrocytes in vivo and antigenic features of oligodendrocyte precursor cells in vitro. This cell category could not be further identified by expression of Ng2 antigen 
, since none of the antibodies tested were working on the rabbit tissue (see ). The absence of Map5, Olig2, and Sox2 immunoreactivity in bipolar-shaped cells does suggest that the newlyborn multipolar cells come from local progenitors in each cerebellar layer.
Antibodies used in this study.
None of the newly generated cells were clearly double-stained for the glial markers GFAP or O4 (). Also double stainings between BrdU and vimentin or S100β, did not yield clear coexpression in identifiable glial cell types, due to the faint expression (S100β) or prevalent localization of these markers in cell processes (vimentin) rather than in the cell body (not shown).
Residual granule cell precursors
Bipolar-shaped cells in the cerebellar cortex of peripuberal rabbits represent about 30% of all cells detectable with the above-mentioned markers (). As described above, a consistent part of these cells are Pax2+/PSA-NCAM+/DCX+ and belong to the population of newly generated neuronal progenitors (). A subpopulation of the PSA-NCAM+/DCX+, but Pax2-negative bipolar cells were immunoreactive for NeuN 
() and Pax6 
(), two markers for cerebellar granule cell precursors (). The detection of Pax6+ and NeuN+ bipolar cells was restricted to the second and third postnatal months, being absent starting from the fourth month and so forth at all later stages (). In addition, Pax6 immunoreactivity was never observed in neuronal-shaped cells (). This indicates that some granule cell precursors continue to migrate from the SPL to the cortical layers during the early phases after the shift from EGL to SPL. Then, this protracted descent of granule cell precursors is exhausted before puberty. The late granule cell precursors seem not to differentiate into mature granule cells, since no immunoreactivity for the GABAA
receptor subunit α6 
was detectable in newlyborn cells (not shown). In addition, consistent apoptotic cell death was present in both layers of the cerebellar cortex in three month-old animals (), whereas it is significantly reduced in the four month-old rabbits, particularly in the granule cell layer (; summarized in ).
Residual granule cell precursors in the cerebellar cortex of peripuberal rabbits.
The occurrence of a residual descent of granule cell precursors raises questions about the persistence of SPL chains of tangentially-oriented cells. These chains are progressively reduced in number from the second to the fifth month, then disappearing from the cerebellar surface during the sixth month. The early neuronal markers doublecortin (), NeuN (), class III β-tubulin 
, and HuC/D protein (not shown) were detected in subpial chains, thus confirming they are formed by neuroblasts (see also 
). Yet, not all antigens detectable in the SPL were co-expressed in cortical cells (see ). For instance, CRMP-4 although present in newlyborn neurons of the molecular layer was absent in SPL chains (). By contrast, the neuronal cell adhesion molecule TAG-1 
was detectable in a subset of these chains () but not in cortical cells (). Similarly, Map5 (Map1B) which is involved in the control of microtubule organization in both neuronal and glial progenitors 
, was present in all SPL chains but absent in DCX+ newlyborn neurons (). These data, in parallel with the neuroepithelial origin of the Pax2+ interneuronal population (see above), suggest that the transient SPL does contain the last granule cell precursors which mostly fail to reach their final destination in the granule cell layer.
Newly generated cells followed for survival times longer than two weeks
Cell countings of BrdU+ cells extended up to two months survival indicated that most of the cells still alive in the cortex two weeks after their birth, actually survive for at least two months, in contrast with a remarkable drop occurring during the first two weeks 
Newly generated cells 1–2 months after their genesis in the cerebellar cortex of peripuberal rabbits.
Although some newlyborn cells were still detectable in association with PSA-NCAM (), DCX (), and Hu protein () up to 30 days since their birth, after this period most of the BrdU+ nuclei were not identifiable through double stainings with cytoplasmic or membrane-bound markers that could reveal a specific cell morphology (Figure S1B
). Unlike neuronal precursor cells, the Map5+ cells started to express the cytoplasmic marker two weeks after their genesis. Then, the Map5 staining was still detectable until 65 days indicating that these cells maintain their multipolar morphology for months ( and Figure S1B
). Accordingly, also the number of BrdU+/Sox2+ double labelled cells was very high between two weeks and two months after their genesis (Figure S2C
No more GABA staining was detectable in newlyborn cells at long survival times, this neurotransmitter being downregulated as cell differentiation occurs 
. In addition, none of the newly generated cells aged more than one month were immunoreactive for the interneuronal markers calbindin, parvalbumin, calretinin (see ).
Some BrdU/vimentin and BrdU/S100β double stainings (not shown) were observed within the granule cell layer and Bergmann glial cell layer (). Most of these cells were not clearly identifiable as to their morphology. This indicates that also astrocytes could be generated at least within the granule cell layer, and suggests that some Bergmann glial cells might proliferate (see below).
Finally, rare Isolectin B4+ (IlB4) microglial cells (3,16 +/− 1,84%) that had incorporated BrdU were detectable at 65 days post-injection survival (Figure S1D
Thus, it appears that a large amount of newly generated neuronal precursors cannot be visualized beyond the first month after their birth, due to a lack of specific cytoplasmic/membrane-bound markers (Figure S1B
). For these reasons, in order to visualize the newlyborn cells at longer survival times, we carried out an ultrastructural study on BrdU-treated rabbits that were left to survive 60 days after the first and 45 days after the last injection of the marker. After visualization of the BrdU+ nuclei by peroxidase staining, small pieces of the cerebellar molecular layer containing these cells were resin-embedded and processed for electron microscopy. The marked nuclei were easily identified in semithin sections () and subsequently photographed in ultrastructure. Cell types corresponding to the size and shape of synantocytes () and neurons () were found, thus showing that elements belonging to both cell populations are still in place after two months. In two cells out of 8 cells analyzed, the occurrence of rare axo-somatic synaptic contacts was observed ().
Newly generated neuronal and glial cell progenitors in the cerebellum of adult rabbits
The cerebellum of adult rabbits was completely devoid of SPL. By contrast, some Ki67+ nuclei were still detectable in the cortex (). The Ki67+ nuclei were frequently detected as ‘doublets’ (), indicating their feature of daughter cell progeny of a recent mitotic event 
. The occurrence of newly generated cortical cells in these animals was confirmed by detection of BrdU after 15 day survival (). The newlyborn cells were present in both cortical layers, more frequently located close to the Purkinje cell layer, on both sides.
Newly generated cells in the cerebellar cortex of the adult rabbit.
By performing the same immunocytochemical analyses carried out in the peripuberal animals, similar PSA-NCAM+/DCX+ neuronal-shaped cells and Map5+ synantocyte-like cells were detected in the adult cerebellum (). Both cell types showed the same morphology, antigenic features, and layer distribution described in the three-month old cerebellar cortex. Nevertheless statistical analysis revealed that in the adult the PSA-NCAM+/DCX+ and Map5+ cells are equlibrated (ML p
0,790; GL p
0,431) Accordingly, the PSA-NCAM+/DCX+ neuronal cells were immunoreactive for GABA (). By contrast, some differences concerned the absence of PSA-NCAM+/DCX+ cells within the white matter of adult cerebella, and the very rare occurrence of bipolar-shaped cells in the cortex (summarized in and ).
Relationships among transcription factors in newly generated neuronal and glial-like cell populations.
Double stainings with DCX or Map5 and BrdU at 15 day survival after the injection revealed that both cell populations in the adult cerebellar cortex were newly generated (). Similarly to what observed in peripuberal rabbits, the PSA-NCAM+ and DCX+ bipolar and neuronal cells co-expressed the two markers (), and most of them were also immunoreactive for CRMP4 () and the transcription factor Pax2 (). Accordingly, the Map5+ stellate-shaped cells co-expressed Sox2 () and Olig2 (not shown). On the other hand, no Pax6+ cell could be detected in the adult cerebellar cortex.
Countings carried out on adult animals (see ) indicated that a further reduction in the number of newly generated cerebellar progenitor cells does occur with respect to the six-month old rabbit, with a ratio of 10/1 at the age of three years (). About one half of these cells are neuronal progenitors that are mostly localized in the molecular layer (see percentages in ).
Relationships among cell proliferation markers, transcription factors, and BLBP in newly generated cells of the postnatal and adult rabbit cerebellum
In order to get more insights about the origin of the two populations of newly generated cortical cells, cross double stainings among different transcription factors were carried out (). These experiments clearly showed that both Olig2 and Sox2 were never coexpressed with Pax2 (), whereas some Sox2+ nuclei also contained Olig2 (). These results support the hypothesis that the Pax2+ neuronal cell precursors belong to a distinct lineage with respect to the multipolar, glia-like cells.
Since it is known in mice that Sox2 is expressed in the nuclei of Bergmann glia 
we used the marker Brain Lipid Binding Protein (BLBP) to visualize cell bodies of Bergmann glia, in order to further investigate this radial glia-derived cerebellar cell type in rabbits. Double labellings showed that virtually all Bergmann glia cells contain Sox2 () but not Olig2 (). In addition, some large Sox2+ nuclei in the Purkinje cell layer did incorporate BrdU at 1–5 days postinjection survival ( and Figure S1C
), and incorporation of BrdU was also detectable in some Bergmann glia cells (). These latter were able to retain the proliferation marker for at least two months (). Also Olig2+ cells can incorporate BrdU at 1–5 days postinjection survival ().
Proliferating Sox2+ nuclei and Brain Lipid Binding Protein (BLBP)+ Bergmann glia cells were also detectable in the Purkinje cell layer of adult rabbits ().
On the whole, these results indicate that Sox2+ dividing Bergmann glia could generate cell precursors that in the following days will express both Sox2 and Olig2, and then Map5, thus becoming visible as multipolar cells (summarized in ). Furthermore, they confirm that Pax2+ neuronal precursors and multipolar glial-like cell precursors generated in the peripuberal and adult rabbit cerebellar cortex belong to two distinct populations.