The base of each fissure is fixed while the lobes grow outward
If initiation of foliation is dependent on Pc anchor points, then the base of each fissure should be fixed relative to the core of the Cb and folia should bulge out in between the anchor points. To address whether this is the case, we superimposed tracings of the outline of midsagittal sections of the mouse vermis at stages between embryonic day (E)16.5 and postnatal day (P)21 (see Materials and methods). In outbred Swiss Webster (SW) mouse embryos, the outer surface of the mouse cerebellar vermis primordium was found to be smooth at E16.5 (Figure 1a). By E17.5, however, the surface was already divided by three slight indentations representing three of the four principal fissures: preculminate, primary, and secondary (Figure 1b, asterisks). At E18.5, the fourth principal fissure (posterolateral) also was evident and the other fissures were deeper (Figure 1c, asterisks). The five cardinal lobes seen in all mammals, and that form between the principal fissures, were thus apparent at E18.5. From anterior to posterior, they are called the anterobasal, anterodorsal, central, posterior and inferior lobes (Figure 1) [9].
Additional fissures (non-principal) successively divided the cardinal lobes into lobules until P7 in SW mice. The anterobasal lobe was the first to be divided by a non-principal fissure at P0 in SW mice, giving rise to lobules I/II and III [8, 17]. In some strains of mice an additional shallow fissure subsequently demarcates lobule I as separate from lobule II, but this was rarely seen in SW mice. In most mouse strains, an additional partial fissure forms in the anterodorsal lobe to partially divide lobules IV and V by a shallow fissure, and this was observed by P5 in the majority of SW mice examined (data not shown). The first subdivision of the central lobe was seen at P1 when the prepyramidal fissure demarcates lobules VII and VIII (data not shown). By P3 the central lobe was further subdivided by the posterior superior fissure into lobules VI and VII (data not shown). In most strains lobule VI is further subdivided by a fissure to form sublobules VIa and VIb, which was seen at P5 in SW mice (data not shown). The posterior lobe is subdivided in some mouse strains to form sublobules IXa, IXb, and IXc. In SW mice the fissure producing IXb and IXc was first apparent at P3, whereas the fissure between IXa and IXb was never seen (data not shown). The inferior lobe is not further divided in mice and is referred to as lobule X (Figure 1d).
By superimposing images of the Cb surface at successive times during the foliation process, we found that indeed the base of each fissure remained in a relatively fixed position and the folia grew outward (Figure 1e). There was a slight outward shift in the positions of the bases of the fissures after P7, likely as a result of the expansion of the white matter and cortex, especially in the central and posterior lobe. Our results are consistent with the base of the fissures functioning as anchors for Cb foliation.
Folding of the Purkinje cell layer predicts the positions of fissures
Since it has been suggested that Pcs, the largest cells in the Cb and sole output of the cortex, are responsible for anchoring the base of the principal fissures, we reasoned that changes in the organization and/or morphology of Pcs might precede formation of fissures [9]. To visualize Pc morphology during initiation of foliation, we stained sagittal Cb sections for the Pc marker Calbindin before (E16.5) and during the initiation of foliation (E17.5 and E18.5) [18]. Anti-Calbindin immunostaining detected a diffuse multilayer of similarly shaped Pcs along the AP axis at E16.5 (Figure 2a). Strikingly, at E16.5 the Pc multilayer was slightly folded inwards (Figure 2a, d, asterisks) at the sites where the three principal fissures formed at E17.5 (Figure 2b, e). By E18.5 all four principal fissures were visible (Figure 2c, f). Interestingly, the invaginations of the Pc multilayer were complementary to sites of inward accumulation of gcps in the external granular layer (EGL) detected by anti-Pax6 immunostaining (insets in Figure 2d–f).
Since Pcs around the base of the fissures at P18 have been found to orient their dendrites towards the base of the fissure, we were curious whether the morphology of individual Pcs was distinct where the fissures emerge at E16.5-E18.5 [19]. We utilized the CreER/loxP based genetic inducible fate mapping (GIFM) approach to mark cells and visualize the cell body and its processes [20]. We used a R26-CreER line (Y Cheng and AL Joyner, in preparation) in combination with a reporter allele that expresses enhanced yellow fluorescent protein (eYFP; R26-eYFP), and induced Cre activity in all cerebellar cell types by administering tamoxifen at E12.5 (Figure 2) [21]. At E16.5 and E17.5, marked cells (YFP positive) were detected in the EGL, as well as in deeper layers of the Cb (Figure 2g–g2, and data not shown). Double immunostaining of cells for anti-green fluorescent protein (GFP) and the Pc nuclear marker anti-RORα identified fate mapped Pcs (Figure 2g–g2) [22]. Within the deeper layer of the Cb cortex, marked cells positive for RORα had small round cell bodies and processes that were extended in various directions (Figure 2g–g2, white arrows). Consistent with previous reports, immature Pcs had numerous randomly oriented projections, and furthermore, Pcs in the emerging fissures had the same morphology with randomly oriented processes (Figure 2g–g2) [9, 23, 24]. In conclusion, the folding of the Pc multilayer with associated accumulation of gcps in the EGL, rather than a change in individual Pc morphology, precedes fissure formation at the outer surface of the Cb.
Purkinje cell maturation is synchronized with fissure lengthening
Since a majority of the maturation of Cb cells occurs during the period of foliation, this raises the question of whether the two processes are linked. In the rat Cb, Pcs located in the anterior (I-V) and posterior (IX-X) folia have been found to mature earlier than Pcs in the central folia [25]. Consistent with the timing being similar in mouse, Pcs in the anterior and posterior lobules of the mouse vermis also are more mature at P7 [26]. To determine whether Pc development proceeds in synchrony with fissure outgrowth, we constructed a developmental profile of the Pcs in the principal fissures compared to later forming non-principal fissures from P0 to P21, using anti-Calbindin immunostaining to visualize Pcs (Figure 3, and data not shown). We judged Pc maturation based on an increase in cell body size and development of dendrites. Initially, the single apical stem dendrite of each Pc appears as an apical swelling, and then the dendrite extends outward and forms a large number of mature dendritic branches [9]. At P0, with the exception of the Pcs in the central lobe, Pcs in the vermis had small cell bodies with an apical swelling and no obvious dendrites, and were organized in a two to three cell thick layer. In contrast, Pcs in the central lobe lacked an apical swelling and were organized into a four to six cell thick layer (Figure 3a–a3). There was no obvious difference in maturation of Pcs at the base versus elsewhere in the forming principal fissures. By P3, Pcs throughout the principal fissures were arranged in a monolayer (Figure 3b1, b3, and data not shown), and the cell bodies had grown in size and a single apical stem dendrite had formed (Figure 3, arrows in 3b1, b3). Curiously, Pcs at the base and the sides of the secondary fissure, which is one of the four principal fissures (between lobules VIII and IX) appeared more mature than elsewhere as they had the largest cell bodies and had begun to form primary dendritic branches (Figure 3b3, arrows).
In contrast to Pcs in the principal fissures at P3, the Pcs in the posterior-superior non-principal fissure, which begins to form at P2, remained in a multilayer, with small cell bodies and no apical dendrite (Figure 3b2). At P5, anti-Calbindin immunostaining was more uniform and stronger in all fissures, perhaps reflecting a progression in Pc differentiation (Figure 3c–c3). In all principal fissures at this stage (Figure 3c1, c3), the Pc bodies had increased in size and primary dendritic branches were evident. Similar to P3, the most mature Pcs were found in the secondary fissure as they had elaborate secondary and tertiary dendritic branches (Figure 3c3, arrow). In the less mature posterior superior fissure at P5, a Pc monolayer had formed (Figure 3c2). At P10, Pcs in the posterior superior fissure had extended and increased their number of secondary and tertiary dendritic branches (Figure 3d2). Pcs in all fissures continued to elaborate the number of their secondary and tertiary dendritic branches after P10 (Figure 3d–d3, and data not shown). In summary, development of Pcs throughout a given fissure proceeds in synchrony with the maturation of the fissure, although Pcs within the secondary fissure are more developmentally advanced from P0 until about P10 than in other principal fissures, suggesting there is no causative link between Pc maturation and formation of fissures.
Granule cell precursors in emerging fissures have a shorter mitotic index than other granule cell precursors and accumulate as inward invaginations
Unlike the Pcs and Bg that are born in the ventricular layer and then migrate radially to form the Pc layer, when gcps leave the ventricular layer (rhombic lip) between E12.5 and E15.5 in mice they migrate over the surface of the Cb to form the EGL and continue proliferating until about P16 [27]. Once gcs become postmitotic, they migrate through the molecular layer and past the Pc layer to form the IGL. A distinct feature of gcps in the EGL that we detected as early as E16.5, one day prior to the first folding of the outer surface of the Cb, was a V-shaped thickening of the EGL, specifically in the positions where the first three principal fissures will emerge (Figure 2a, d). One possible mechanism for the increased number of gcps at these positions is an increase in the rate of proliferation. We therefore examined gcp proliferation at the onset of Cb foliation using markers for different stages of the cell cycle. To mark gcps in the DNA replication phase (S phase) at E17.5 and E18.5, bromodeoxyuridine (BrdU) was administered approximately 20 minutes prior to analysis of the embryonic Cb (Figure 4a, b). This resulted in marking of approximately 26% of cells within the EGL and did not reveal an obvious difference in the number of cells incorporating BrdU at the bottom or the crown of the folia at E17.5 or E18.5 (Figure 4a, b).
In order to mark a shorter phase of the cell cycle than S phase, nuclei undergoing the G2/mitosis phase were marked using anti- phosphohistone 3 immunostaining. In contrast to the seemingly uniform BrdU labeling throughout the AP axis of the EGL, quantification of pH3 positive gcps (Figure 4c) revealed that pH3 positive gcps were more frequently found at the sites where the first three principal fissures were emerging then elsewhere at E16.5 and E17.5 (Figure 4c, d, arrows; data not shown). However, this was not the case for these fissures at E18.5 when pH3 positive gcps did not reveal a significant difference in the distribution throughout the EGL (Figure 4c, e, arrows; data not shown). This result suggests that gcps in the area of the emerging fissures transiently divide more often than other gcps during initiation of fissure formation due to a shorter cell cycle. Furthermore, possibly because there is less resistance in the Cb cortex than in the overlying basal lamina, the gcps invaginate inwards to produce the first morphological manifestation of the base of the fissures. By E18.5, the width of EGL at the base was only slightly thicker than at the crown of principal fissures, consistent with our observation that cells in the base of the fissures no longer had an obviously higher mitotic index (Figures 2f and 4b). Moreover, during establishment of the non-principal fissures at later stages, gcps were found to accumulate where the fissures later formed, indicating that this process is conserved for all cerebellar fissures (data not shown).
Granule cell precursor morphology and organization is distinct at the base of emerging fissures
Cell shape changes have been implicated as a driving force in formation of grooves or invaginations in many model systems [28]. Since there are no obvious differences in the shape or orientation of Pcs at the base of the fissure between E16.5 to E18.5 (Figure 2), we asked whether changes in gcp morphology accompany the initial formation of fissures (Figure 5). We characterized the shape of the gcp cell bodies by examining GFP expression in transgenic mice that ubiquitously express a cell membrane localized GFP (CAG::GPI-eGFP) [29]. To quantify differences in cell morphology, we determined the circularity index (ci; Figure 5e) of GFP+ gcps in the first three emerging fissures (within the invagination) at E16.5 (Figure 5a, a1), E17.5 (Figure 5b–b2) and P0 (Figure 5c, d) compared to gcps between the fissures (at the crown) [30]. The EGL was identified as the outer most five cell layers at the crowns of the folia and seven cell layers at the base of the fissures by double immunostaining for anti-GFP and anti-Pax6, a gcp marker or anti-RORα, a Pc marker (Figure 5a1, b, c). At E16.5 when the outer surface of the Cb is smooth, proliferating gcps throughout the AP axis of the midline vermis had a round shape (Figure 5a2, e; ci = 0.830 at the crown of folia, ci = 0.8 at the future base of fissures) [31]. At E17.5 when the outer surface begins to invaginate, gcps in these emerging fissures were significantly less round (Figure 5b1, e; ci = 0.685) than gcps at the midpoint between fissures (Figure 5b2, e; ci = 0.810, p < 0.0001). Furthermore, at E17.5 the axis of elongation of gcs at the base of the fissures was uniformly parallel to the fissure. In contrast, there was no preferential orientation of the longitudinal axes of gcps at the crown of folia (Figure 5b2, and data not shown). At P0, the difference in gcp shape between the bases of the three principal fissures (Figure 5c, e; ci = 0.646) and the crowns (Figure 5d, e; ci = 0.837) was even more evident (p < 0.0001). Furthermore, we found that at E17.5 and P0 elongated gcps are mitotic since they were positive for anti-pH3 marker (Figure 5d, and data not shown).
To further characterize gcps in emerging fissures and to confirm our findings in CAG::GPI-eGFP transgenic mice, we used electron microscopy to analyze sagittal sections at E17.5 (Figure 5f). The gcps located at the base of the fissure were clearly elongated (ci = 0.65) and their longitudinal axes were parallel to the fissure (Figure 5f). In contrast, gcps located five to seven cells away from the base of the fissure were more round (ci = 0.78) without any consistent alignment of their axes (Figure 5f). Interestingly, gcps found in between these two positions had an intermediate phenotype in terms of cell shape (ci = 0.76), but they lacked an organized orientation of their longitudinal axes. Thus, gcp cell body elongation and an inward accumulation of gcps due to an increased proliferation are clear hallmarks of the emergence of the fissures and these cellular changes may drive the inward folding of the Pc layer. We therefore define the entire region undergoing these unique morphogenetic changes as an 'anchoring center' for each fissure.
The onset of granule cell differentiation is concomitant with cerebellar foliation
We next explored whether the migration and differentiation of gcps from the EGL to the IGL is different in fissures from the rest of the folia. Gcps proliferate only in the outer EGL (oEGL) and then move to the inner EGL (iEGL) when they begin to differentiate. Postmitotic gcs within the iEGL extend parallel fibers along the medial/lateral axis and undergo nuclear translocation along one parallel fiber before extending a radial process and descending along Bg fibers past the Pcs to form the IGL [9, 32, 33]. The gc parallel fibers constitute part of the molecular layer. Anti-p27/Kip1 and anti-NeuN were used to mark differentiating gcs both in the iEGL and during their migration to the IGL (Additional file 1). The earliest a diffuse IGL layer could be detected using these markers was at P1 (Additional file 1a, a1). As has been reported, we found the IGL to be thinner at the base of the fissures than the sides and thickest at the crown of the folia at all stages analyzed between P3 and P21, but not at P1 (Additional file 1b–b2, c–c2, and data not shown) [9].
Since the lack of gc differentiation before P1 (that is, during emergence of the principal fissures) could be due to lack of expression of p27/NeuN rather than an absence of differentiation of gcs, we utilized GIFM to mark postmitotic gcs independent of the timing of expression of gc marker genes. Given that all gcps express the transcription factor Math1, we used a Math1-CreER transgenic line in combination with a postmitotic neuron reporter allele (Tau-lox-STOP-myrGFP-nlacZ), and induced Cre activity in embryonic gcp by administering tamoxifen at E15.5 (Figure 6) [34, 35]. The Tau reporter allele initiates permanent expression of membrane localized GFP and nuclear lacZ (nlacZ) in neurons as they begin to differentiate. Coronal sections at E17.5 revealed marked gcs (GFP/nlacZ double positive) both in the iEGL, based on their position below the outer three to five cell layers of the EGL and labeling with Semaphorin6a (Sema6a), a marker for gcs undergoing nuclear tangential migration (Figure 6a, b, e–e3), as well as in deeper positions (Figure 6a, b, e-e3) [36, 37]. Double immunohistochemistry for anti-Pax6 and anti-GFP on sagittal sections at E18.5 confirmed that all the marked cells were indeed gcs (Figure 6c). The combination of nlacZ and myrGFP immunostaining revealed that the gcs within the iEGL had a round cell body and parallel fibers (Figure 6a, arrow). Within the deeper layer of the Cb, below the iEGL, some gcs had elongated cell bodies parallel to the surface of the Cb and a descending radial fiber (Figure 6b, arrow). Furthermore, we found no obvious differences in morphology and orientation of the gc bodies, or their parallel fibers at different AP positions, suggesting that the process of gc differentiation is similar at the base of fissures and crowns of folia. By E18.5 descendents of the gcps marked at E16.5 could be seen accumulating in a broad and diffuse layer below the Pcs (anti-Calbindin positive cells; Figure 6d), consistent with our detection of a sparse EGL at P1 using anti-p27/Kip1 or NeuN (Additional file 1a, and data not shown).
To address whether gcps differentiation occurs at the same time at the base of the fissures versus the crown of the folia, we used GIFM to mark gcps with tamoxifen at E15.5 followed by a BrdU pulse at E16.5, and analysis of midsagittal Cb sections at E18.5 (Figure 6f–f2). Quantification of cells labeled for anti-βgalactosidase (anti-βgal) in the base of the first three fissures to form versus the crown of the adjacent folia revealed that gcs differentiate at both positions at E16.5 (Figure 6g). Interestingly, we found more differentiated gcs at the base of the fissures (mean = 8.18 cells below a 10 micron region of the iEGL; Figure 6g) than at the crowns of the folia (mean = 6 cells below a 10 micron region of the iEGL; p < 0.006; Figure 6g). Quantification of anti-βgal and anti-BrdU double positive gcs showed a tendency toward an increase in the number of differentiated gcs born at E16.5 at the base of the fissures (mean = 2.85 cells below a 10 micron region of the iEGL; Figure 6g) versus the crown of the folia (mean = 2.22 cells below a 10 micron region of the iEGL; p < 0.05; Figure 6g). Moreover, we found double positive (BrdU/βgal) gcs only in the most medial sections of the vermis (within 80–100 μm of the midline), where the three principal fissures are the longest at E18.5. In conclusion, gcs begin their differentiation program as early as E17.5 in the most medial Cb, coinciding with the position where foliation is first observed, and preferentially at the base of the emerging fissures.
Bergmann glial fibers fan out from a single central point at the base of fissures
Each Bg cell projects a single fiber to the pial surface that forms a specialized structure called a glial endfoot that together with the basal lamina maintains the integrity of the Cb [9, 38, 39]. Furthermore, Bg provide a scaffold for Pc dendritic outgrowth and for gc migration [9, 40–43]. Thus, Bg fibers could contribute to the emergence or maturation of fissures. To investigate whether Bg play a such role in anchoring centers, we used anti-RC2, anti-brain lipid binding protein (anti-BLBP) and anti-glial fibrillary acidic protein (anti-GFAP) immunostaining to visualize the orientation of the Bg fibers from E16.5 to P14 (Figure 7, and data not shown) [44–47].
In order to establish the orientation of the Bg fibers in relation to the first manifestations of anchoring centers (accumulation of gcps and invagination of the Pc layer), we double labeled for Bg and gcps or Pcs. To label gcps, we utilized a 20 minute BrdU pulse to identify the proliferative layer of the EGL (anti-Pax6 positive in control experiments; see Additional file 2). At E16.5 and E17.5 anti-RC2 immunostaining revealed that Bg fibers were oriented parallel to each other and perpendicular to the Cb outer surface, even in the emerging fissures where gcps accumulated and the Pc layer invaginated (Figure 7a–j). Interestingly, at E18.5 (Figure 7k–s), anti-BLBP immunostaining revealed that the Bg surrounding the emerging principal fissures projected their fibers to a single point at the base of the fissure (Figure 7l, asterisk). In contrast, the remaining Bg fibers were oriented nearly parallel to each other and aligned perpendicular to the pial surface of the Cb (Figure 7p). At later stages, this specific organization of Bg fibers at the base and the sides of the fissure was more pronounced (data not shown). Our analysis suggests that the glial endfeet of Bg fibers surrounding the base of the emerging fissures form a hub from which the fibers fan out. This suggests that Bg fibers do not contribute to the initial formation of fissures, but play an important role in the function of the anchoring centers by directing migration of gcs at the base of the fissure in a semicircle. Furthermore, this spreading out of the gcs could account for the thinner IGL at the base of the fissures.
An alteration in the time when two anchoring centers form underlies the altered vermis foliation pattern in Engrailed2 mutants
Our results indicate that a precise coordination of distinct behaviors of Pcs, gcps, and Bg fibers produces anchoring centers that function as the base of each fissure. By extrapolation, the position and timing of formation of each anchoring center should determine the shape and organization of the folia. To test this hypothesis, we examined a mouse mutant with an altered foliation pattern to determine whether a change in the timing and/or positioning of the key cellular events responsible for initiation of fissures occurs in accordance with the foliation defect. We chose to analyze mice lacking the gene encoding the homeobox transcription factor Engrailed2 (En2), since in the vermis of En2 mutants the secondary principal fissure (which normally separates the central lobe from the posterior lobe) is shifted posterior such that lobule VIII of the central lobe is associated with lobule IX of the posterior lobe rather than with lobules VI/VII of the central lobe [48]. In addition, the relative lengths of the secondary and prepyramidal fissures are reversed (Figure 8a, b) [48–52]. The specific and reproducible alteration in only one folium of En2-/- mutants affords the opportunity to study the formation of the two anchoring centers surrounding lobule VIII amidst others that produce normal folia. We therefore examined whether the vermis foliation phenotype in En2 mutants is associated with a change in the timing and/or positioning of the secondary and prepyramidal anchoring centers that surround lobule VIII.
At E17.5, the overall size of the En2 mutant Cb was reduced, the surface of the Cb smooth rather than having three principal fissures, and the EGL thinner than normal (Figure 8c, d), consistent with our previously documented general postnatal delay in foliation and an overall reduction in the size of the mutant adult Cb [50]. Furthermore, we found that at E17.5 the Pc layer of En2 mutants lacked invaginations (data not shown) and there were no signs of accumulation of gcps. An observed weak Calbindin immunostaining in En2 mutants might further reflect a general developmental delay in Pc maturation (data not shown). Interestingly, by P0 all the principal fissures except the secondary fissure were clearly visible in En2 mutants (Figure 8e, f). The Pc layer had, however, begun to fold and gcps accumulated inwards where the secondary fissure was expected to form. Moreover, despite the overall delay in formation of fissures in En2 mutants, the Pc layer had already begun to invaginate and gcps to accumulate at P0 in the area where the prepyramidal fissure should form, a few hours earlier than in wild-type mice (Figure 8g, h). At P1 the prepyramidal fissure of En2 mutants was deeper than the newly forming secondary fissure, whereas in wild-type mice the secondary fissure is much deeper than the prepyramidal fissure (Figure 8i, j). Furthermore, the Pcs found in the secondary fissure in En2 mutants at P1 were smaller in size and less mature than normal, whereas Pcs found in the prepyramidal fissure were more mature than Pcs found in the wild-type prepyramidal fissure (Figure 8i, j). By P2, the prepyramidal fissure had continued to lengthen in En2 mutants more than in wild-type mice, and the secondary fissure remained shorter than normal (Figure 8m, n). Additionally, altered timing of the organization of Bg fibers correlated with the altered fissure formation in En2 mutants (Figure 8k, l, and data not shown). It was not until P1 in En2 mutants when the prepyramidal and secondary fissures had formed that the Bg fibers fanned out from the two anchoring centers (Figure 8k, l). Consistent with the apparent altered timing of formation of the prepyramidal and secondary anchoring centers in En2 mutants, at P0 and P1 the gcps in the emerging secondary fissure of En2 mutants were elongated more than gcps between the fissures, but slightly less than normal (wild type: ci = 0.55 at P0 ci = 0.52 at P1 (Figure 9a, i, and data not shown); En2 mutant: ci = 0.59 at P0, ci = 0.59 at P1 (Figure 9c, i, and data not shown) and the gcps in the anchoring center of the mutant prepyramidal fissure were more elongated (ci = 0.59 at P0, ci = 0.5 at P1; Figure 9g, i) than normal (ci = 0.71 at P0, ci = 0.62 at P1, p < 0.0009; Figure 9e, i, and data not shown). Moreover, at P0 in En2 mutants quantification of pH3 positive gcps (Figure 9j) revealed that pH3 positive gcps were more frequently found in the emerging secondary and prepyramidal fissure then in the crown of the lobule between these two fissures. Our results demonstrate that in the prepyramidal and secondary anchoring centers of En2 mutants the normal procession of morphogenetic behaviors of Pcs, gcps and Bg fibers occurs in unison with the premature formation of the prepyramidal fissure and delayed formation of the secondary fissure compared to the same wild-type fissures. This altered timing of anchoring center formation then leads to the apparent posterior shift of lobule VIII.