Common Origin of the Cerebellar Dual Somatotopic Areas Revealed by Tracking Embryonic Purkinje Cell Clusters with Birthdate Tagging

Visual Abstract


Introduction
Representation of the somatotopy is deeply involved in the motor control function in the cerebellum (Manni and Petrosini, 2004). Electrophysiological and neuroimaging studies have shown dual somatotopic areas in the anterior and posterior cerebellar lobules which is one of the noticeable features of the functional localization of the cerebellar cortex of humans and other mammals ( Fig. 1A; Thickbroom et al., 2003;Manni and Petrosini, 2004).
Morphologically, the cerebellar cortex is organized by multiple longitudinal striped subdivisions. Two types of mutually linked subdivisions, 1) modules and 2) molecular compartments, have been identified. The modules have been defined by the topographic connections of PC axons and climbing fiber axons ( Fig. 1B; Voogd and Glickstein, 1998;Apps and Hawkes, 2009;Cerminara et al., 2013;Fujita and Sugihara, 2013;Ruigrok et al., 2015), whereas the molecular compartments have been defined by the arrangement of Purkinje cells (PCs) that show heterogeneous expression of marker molecules such as zebrin II or aldolase C ( Fig. 1C; Brochu et al., 1990;Voogd and Glickstein, 1998;Sugihara and Shinoda, 2004;Sillitoe and Joyner, 2007;Fujita et al., 2014). The somatotopic representation is most clearly seen in the paravermal area in anterior and posterior lobules, in which zebrin-negative and -faintly-positive stripes (identical to the C1/C3 module; Fig. 1B, C) occupy substantial proportions of the cerebellar cortex. Both the anterior and posterior parts of this area are topographically innervated by the climbing fiber axons originating from the dorsal accessory olive and project to the anterior interposed nucleus (Ekerot et al., 1997;Cerminara et al., 2013;Ruigrok et al., 2015;Low et al., 2018) to be involved in the control of fine body movements such as grasping and limb cutaneous reflexes in the cat (Horn et al., 2010) and rat (Pijpers et al., 2008). Because the C1/C3 module represents the main part of the cerebellar somatotopic area as mentioned above, the anteroposterior separation of the C1/C3 module (Fig. 1B) may be the anatomical correlate for the anteroposterior dual representation of somatotopy observed in animal and human cerebellums (Snider, 1950;Stoodley et al., 2012;Guell et al., 2018).
PCs are born in the period between embryonic day 10.5 (E10.5) and E12.5 in the ventricular zone (Hashimoto and Mikoshiba, 2003) and form the main body of the immature cerebellum by E14.5 in mice (Goffinet, 1983). At E14.5, some eight heterogeneous subsets of PCs are arranged in clustered compartments as observed by molecular marker labeling (Vibulyaseck et al., 2017) or genetic profiling (Wizeman et al., 2019). At E17.5, the number of heterogeneous populations of PCs increases to about 50, which are arranged into clusters separated by PC-free gaps ( Fig. 1E; Korneliussen, 1968;Altman and Bayer, 1985;Smeyne et al., 1991;Oberdick et al., 1993;Millen et al., 1995;Larouche et al., 2006;Wilson et al., 2011;Fujita et al., 2012;Carter et al., 2018;Wizeman et al., 2019). Each of the E17.5 clusters develops directly into an individual adult PC stripe in the postnatal period ( Fig. 1D; Sillitoe et al., 2009;Namba et al., 2011;Fujita et al., 2012). Therefore, we hypothesized that the rearrangement of embryonic PC clusters is essential in shaping the compartmental organization of the adult cerebellar cortex which includes its modular organization and the dual somatotopic areas. However, accurate spatial tracking of lineages of all E14.5 clusters would be required to test this hypothesis.
Each striped compartment in the adult cerebellar cortex contains PCs generated at particular timing (Hashimoto and Mikoshiba, 2003;Namba et al., 2011;Zhang et al., 2020). Therefore, birthdate-specific labeling of PCs can be a useful technique to track the cerebellar compartmentalization. The CreER-LoxP system that targets the ascl1 gene, which is transiently expressed at the time of neuronal differentiation, can label neurons that are born at the time of tamoxifen injection (Sudarov et al., 2011). We used a similar birthdate-tagging system (G2A mouse line, Hirata et al., 2019;Zhang et al., 2020) that targets the Neurog2 gene, which is expressed in neurons including PCs (Zordan et al., 2008), when the neuronal progenitors start differentiating (Florio et al., 2012).
By combining birthdate-specific labeling and molecular marker labeling of PCs (Minaki et al., 2008;Fujita et al., 2012), we tracked the migration and division of all embryonic PC clusters from E14.5 to E17.5 to clarify the spatial development of the cerebellar compartmentalization. We then focused on the lineage of a particular E14.5 cluster, the fate of which was crucial to test the above hypothesis.

Histological procedures
A Caesarean section was performed on pregnant Ai9 females anesthetized with an intramuscular injection of pentobarbital sodium (0.1 mg/g body weight) and xylazine (0.005 mg/g body weight) to obtain E14.5-E17.5 embryos at noon. Embryos were perfused transcardially with phosphate-buffered saline (PBS, pH 7.4) with heparin sulfate (0.1%), and then with 4% paraformaldehyde. The anesthetized female was sacrificed by cervical dislocation after removing embryos. The embryo brains were dissected in chilled 4% paraformaldehyde and kept in 4% paraformaldehyde for post-fixation and then soaked in sucrose solution (30% with 0.05M phosphate buffer, pH 7.4) for two days. Among embryo samples obtained from pregnant Ai9 female, ones that showed patterned red fluorescence reporter expression in the brain were regarded as G2A::Ai9 heterozygous hybrid. Brain samples were stored in the deep freezer until sectioning.
G2A::Ai9::AldocV heterozygous hybrid mice were anesthetized with an intramuscular injection of an overdose of pentobarbital sodium (0.18 mg/g body weight) and xylazine (0.009 mg/g body weight) at postnatal day 42. They were perfused transcardially with PBS with 0.1% heparin sulfate, and then with 4% paraformaldehyde. The skull was kept in 4% paraformaldehyde for post-fixation overnight. The brain was dissected and soaked in sucrose solution for two days. Brain samples were stored in the deep freezer until sectioning.
Embryo G2A::Ai9 brains were coated with gelatin solution (10% gelatin, 10% sucrose in 10mM phosphate buffer, pH 7.4, 32°C). The gelatin block was hardened by chilling and then soaked overnight in fixative with a high sucrose content (4% paraformaldehyde, 30% sucrose in 0.05 M phosphate buffer, pH 7.4). Complete sets of serial sections were cut coronally, horizontally and sagittally using freezing microtome at a thickness of 40 mm. The ventral surface of the medulla was regarded as the horizontal plane. After washing in PBS and PBS with 0.12% Triton X-100 (PBST), each complete set of sections was processed for immunostaining. Floating sections were incubated on a shaker with a mixture of two or three primary antibodies produced in different host animal species in PBST plus 2% normal donkey serum for 48 hours at 4°C . Goat anti-EphA4 (R&D Systems), goat anti-FoxP2 (Everest Biotech), rabbit anti-Corl2 (provided by Dr. Ono at KAN Research Institute) and rat anti-OL-protocadherin (Millipore) are the primary antibodies used in the majority of experiments. Rabbit anti-FoxP2 antibody (Abgent) was used in combination with the goat primary antibody. The specificity of the above antibodies has been described (Vibulyaseck et al., 2017). In some experiments, mouse anti-Calbindin-D28k (Sigma-Aldrich) and rabbit anti-Calbindin-D28k (AnaSpec) antibodies were also used ( Table 1). The sections were then incubated with a mixture of appropriate two or three secondary antibodies that were conjugated with fluorescent tags (Table 1). Some sections were counterstained with 4', 6-diamidino-2-phenylindole dihydrochloride (DAPI; 1:3,000; D212, Dojindo, Mashiki, Kumamoto, Japan). Finally, these sections were mounted on glass slides, dried, coverslipped with watersoluble mounting medium (CC mount, Sigma C9368-30ML).
Postnatal G2A::Ai9::AldocV brains were embedded in gelatin, and cut coronally into serial sections at a thickness of 50 mm. The complete sets of sections were mounted on glass slides, dried, coverslipped with watersoluble mounting medium (CC mount).

Acquisition of digital images
Fluorescent images were digitized using a cooled color CCD camera (AxioCam1Cm1, Zeiss, Oberkochen, Germany) attached to a fluorescent microscope (AxioImager.Z2, Zeiss) in 12-bit gray-scale with appropriate filter sets. To digitize a section of the cerebellum, 2.5X objective and tiling function of the software to control digitizing (Zen 2 Pro, Zeiss) was used. Images of all serial sections of a brain were obtained with the same exposure and gain parameters. Images were adjusted in contrast and brightness and assembled using a software (ZEN 2 Pro, Zeiss and Photoshop 7, Adobe, San Jose, CA, USA). High magnification confocal images were taken with a 63X objective lens and appropriate filters and laser light sources attached to the confocal microscope (TCS SP8, Leica, Wetzlar, Germany). Images were adjusted in contrast and brightness and assembled using software (Las X, Leica). A combination of pseudo-colors was applied to fluorescent and confocal images in figures. Photographs were assembled using Photoshop and Illustrator software (Adobe). Digital enhancements were applied to entire images and no manipulations were applied other than contrast or brightness.

Three-dimensional reconstruction of Purkinje cell clusters
Three-dimensional (3D) models of PC clusters were reconstructed through two steps: 1) two dimensional (2D) drawings of contours of identified PC clusters, and 2) 3D surface modeling with these 2D drawings. Digital images of serial sections were placed in individual layers of 2D graphic software (Adobe Illustrator 10). Their positions and orientations were then adjusted by superimposing them on each other while referring to landmark structures such as the midline, the cerebellar surface and major labeled areas. The cerebellar surface and contour of groups of Purkinje cell subsets and marker-labeled areas (i.e. PC cluster) were drawn using curve tools of Illustrator in all sections. Distribution patterns of PCs and PC-free gaps, and expression patterns of PC markers were systematically observed in this procedure. Cerebellar nuclear areas identified by the lack of Corl2 signal were excluded from the reconstruction. After the identification of PC clusters, all drawings in sections of coronal, horizontal and sagittal planes were imported into 3D graphics software (Rhinoceros 4, Robert McNeel & Associates, Seattle, WA, USA), with the z-axis position adjusted for each section. The 3D drawings of cerebellar structures obtained from brain samples with different cutting orientations were matched and compared with one another to identify structures. The 3D surface reconstructions were made from a set of coronal section drawings by using the 'loft' command in Rhinoceros (Fujita et al., 2012). Cerebellar fissures were reconstructed in the 3D space from aligned drawings from sagittal sections.
Definition of the relative position of a coronal plane within the cerebellum The relative position of a section in the whole extent of the cerebellum was defined using percentages as described previously (Vibulyaseck et al., 2017). In short, the position of the most caudal section of the coronal sections in which the cerebellum first appeared was defined as 0%, whereas the position of the most rostral section in which the cerebellum remained was defined as 100%. The position of other sections was obtained by linear interpolation. In the case of horizontal sections, the most dorsal section was defined as 0%, while the most ventral section in which the cerebellum remained was defined as 100%.
At E14.5, nine PC clusters-termed medial, dorsal, central-medial, central-lateral, mid-lateral, lateral, dorsolateral, rostrodorsolateral, and ventrolateral (m, d, c-m, c-l, ml, l, dl, rdl, and vl;based on Vibulyaseck et al., 2017)-were arranged in column-shaped elongations in the rostrocaudal direction at various mediolateral and dorsoventral levels ( Fig. 2I). A previously designated Pcdh10-positive cluster, termed "c" (Vibulyaseck et al., 2017), was revised here: it was divided into the central-medial (c-m) and central-lateral (c-l) clusters because the expression of Corl2 was stronger in the c-m than in the c-l ( Fig. 2A-C). The distribution of these nine clusters was remarkably consistent with that of the unbiasedly classified groups of PCs via single-cell RNAseq (Wizeman et al., 2019;see Discussion). These clusters, thus, may reflect fundamental molecular distinctions in PCs at E14.5.
The number of PC clusters increased from 9 at E14.5 (Fig. 2), to 18 at E15.5 (Fig. 3), and to 28 at E16.5 (Fig. 4). The number increased presumably because 1) new PCfree gaps appeared inside a cluster, and/or 2) a part of a cluster changed in molecular expression from the other part of the same cluster. Furthermore, some divided clusters seemed to migrate separately. At E17.5, PC clusters were narrow in the mediolateral direction but often extended rostrocaudally (longitudinally) across immature lobules, somewhat resembling adult striped compartments, although they were not yet arranged in a single layer but in multiple layers, shallow or deep from the cerebellar surface (Fig. 5;Fujita et al., 2012).
Although E17.5 cerebellum contains 54 clusters identified with detailed analyses (Fujita et al., 2012), this study focused on the more qualitative distinction of clusters to facilitate analyses and simplify description. Namely, neighboring clusters that had only slightly different molecular expression profiles and/or not clearly separated from one another by intercalating PC-free gaps were combined. For example, our cluster vp1-2 includes clusters vp1 and vp2 of Fujita et al. (2012). We combined 11 sets of two or three neighboring E17.5 clusters into single clusters, resulting in a total of 37 clusters in place of 54. We adopted the nomenclature (Table 3) from Fujita et al. (2012) to designate E17.5 clusters in the present study.
Tamoxifen administration labeled neurons with tdTomato in the cerebellum of G2A::Ai9 mice ( Fig. 6A-D). The labeling pattern was dependent on the timing of tamoxifen administration but consistent among cases that had the same administration timing. We first checked the specificity and efficiency of the labeling in the E14.5 cerebellum. All tdTomato-expressing cells (100%) inside a PC cluster coexpressed Corl2 (Fig. 6F), in all mice (n=23), indicating that tdTomato labeling was specific to PCs. Efficiency of the labeling was estimated by counting the number of tdTomato-expressing PCs among all PCs. Because the majority of PCs are born between E10.5 and E12.5, the sum of the ratios of the tdTomato-labeled PCs in the TM10.5, TM11.5, and TM12.5 cerebellums would become close to 100% if the efficiency is high. Indeed, in the 10,000-mm 2 area of dense PC distribution within the ml cluster of TM10.5, TM11.5, and TM12.5 cerebellums (n=1 each) at E14.5, 41.1% (44 PCs out of 107, 44/107), 52.0% (52/100), and 4.7% (5/106) of PCs were tdTomatopositive, respectively, indicating an estimated efficiency of 97.8%. A similar measurement in the c-l cluster showed 1.1% (1/89), 40.8% (40/98) and 45.2% (42/93) labeled PCs in TM10.5, TM11.5 and TM12.5 cerebellums (n=1 each), respectively, at E14.5, indicating an estimated efficiency of 87.1%. The results indicated that the recombination is highly efficient as well as highly dependent on the timing of tamoxifen administration. Similar efficiency and timing-dependency have been also observed in PCs in the adult cerebellum (Zhang et al., 2020). Thus, the PCs labeled by the tamoxifen injection on a specific timing were designated as "E10.5-born" etc. and the area or cluster in which about 1% of PC were labeled with tamoxifen injection at E10.5 was designated as "E10.5-PCsparse".
We then examined the distribution of E10.5-born, E11.5-born and E12.5-born PCs in the PC clusters, which were identified as described in the preceding section, in the E14.5 cerebellum. (Fig. 6A-D and Table 2). E10.5born PCs were observed densely in the d, ml, and rdl clusters, moderately or sparsely in the c-m, l, dl, and vl clusters (Fig. 6A1, B1, C1, D1). However, almost none of the E10.5-born PCs were observed in the m and c-l clusters (blue asterisks in Fig. 6A1, B1, C1). E11.5-born PCs were observed in all clusters (Fig. 6A2, B2, C2, D2). They were more densely distributed in the d, ml, and rdl clusters than in other clusters, and were absent in the medial part of the m cluster. E12.5-born PCs distributed densely in the m cluster and moderately in the c-m, c-l, l, and vl clusters (Fig. 6A3, B3, C3, D3), but rarely contributed to either the d, ml, dl or rdl clusters (orange asterisks in Fig. 6A3, B3, C3, D3). This observation was further quantified by measuring the fluorescence signal intensity, which was supposed to be approximately linearly related to the density of labeled PCs, in each cluster, in digital images of sections of TM10.5, TM11.5 and TM12.5 cerebellums (Fig.  6E). The labeling densities of the c-l and m clusters were near 0, lower than those of other clusters in the TM10.5 cerebellum. but increased to higher levels in the TM11.5 and TM 12.5 cerebellums. On the contrary, the labeling densities of the ml, d, and rdl clusters were higher than those of other clusters in the TM10.5 and TM11.5 cerebellums, but decreased to the level near 0, lower than those of other clusters in the TM12.5 cerebellum (Fig. 6E).
The above conclusion was supported by the consistency of the marker expression profile of lineage clusters. The m, d, c-m, c-l, l, vl lineage clusters showed similar marker expression profiles to those of their original E14.5 clusters, although minor changes in the expression profile were sometimes recognized in some cases (Table 4). Among ml, dl, and rdl lineage clusters, which were E12.5-PC-sparse and located next to one another in the lateral cerebellum, the ml cluster at E14.5 and its daughters, or the ml lineage clusters, at later stages consistently showed strong Pcdh10 expression and were identified consequently (lateral orange in Fig. 7A2, B2, C2, D2). The dl and rdl lineage clusters were distinguished based on denser E10.5-born PCs and Pcdh10 expression and more rostrolateral positioning of rdl lineage clusters than dl lineage clusters (Table 4). Since their distinction was not necessarily very clear, we sometimes indicate them together by "dl1rdl" in this report.
As a whole, the birthdate-specific labeling of PCs allowed us to identify the lineage of all PC clusters between E14.5 and E17.5 (Table 3).
The results demonstrated that the positional relationship among clusters changed in some places during the period between E14.5 and E17.5, indicating that the separate migration of divided clusters is one of the bases for the rearrangement of the compartmental organization of the cerebellum. Furthermore, the marker expression profiles of separated clusters that belonged to the same lineage changed differently to some extent in some cases (Table 4).
Spatial differentiation of the c-l lineage clusters during the period from E14.5 to E17.5 Among the 37 identified clusters in the E17.5 cerebellum, c-l lineage clusters were located in the most widelyseparated areas (light blue, ip1-2, ia1-2, va2-4, it3, ia3-5 and ha4 in Fig. 7D4). Although originating from a single c-l cluster, the c-l lineage clusters were located separately at various rostrocaudal and mediolateral positions in the E17.5 cerebellum (Fig. 8G). As described in the preceding section, all the c-l lineage clusters were E10.5-PC-sparse, which facilitated their identification, and showed no expression of Pcdh10 (scarce red or green signals in   Fig. 8A-F) and various expression of EphA4 (blue signals in some circumscribed clusters in Fig. 8A-F).
To further clarify the spatial differentiation of c-l lineage clusters during the period between E14.5 and E17.5, we examined positional relationships between c-l lineage clusters and neighboring clusters (Fig. 9). At E14.5, the c-l cluster was located lateral to the c-m cluster and medial to the ml cluster: again, the c-l cluster was identified by the weak Pcdh10 expression, weak Corl2 expression and the lack of E10.5-born PCs (Fig. 9A1, Table 2). It occupied the mid-lateral part of the center of the cerebellum with its dorsocaudal part extended laterally to the superficial area dorsal to the ml cluster (Fig. 9A1, B1). The rostral part of the c-l cluster was more extended laterally in the position rostroventral to the ml cluster (arrowheads in Fig.  9A1). The rostromedial part of the c-l cluster had a slightly stronger Corl2 expression than the rest of the c-l cluster (Fig. 9A1, asterisk) and adjoined medially with the c-m cluster which had strong Pcdh10 expression (Fig. 9C1,D1). continued (green) and immunostaining signals of Pcd10 (red) and EphA4 (blue) are shown from the top to the bottom. Blue and orange colors in dashed lines (in subpanels in 1 and 2), circumscribing lines (in subpanels 3), and clusters in the 3D schemes (in subpanels 4) indicate E10.5-PC-sparse and E12.5-PC-sparse clusters. Black line in the 3D scheme indicates the position of the coronal section in subpanels 1-3. Abbreviations,d,dl,l,m,ml,rdl,vl,names of E14.5 clusters;ha4,ha6,hc1,hc2,hp4,ia4,ic3,it2,it3,va1,vc1,vt1,vt2,vt3,vt4,vt5,names of E17.5 clusters;C,caudal;D,dorsal;L,lateral;M,medial;R,rostral.V,ventral. Fujita et al. (2012), and location of the cluster within the caudorostral level of the cerebellum are shown (columns 2,3). Relative intensity in immunostaining in the present study (columns 4-8) was generally consistent with the previous result (Fujita et al., 2012), but showed some minor differences. In cluster names, "v", "i", "h", "a", "p", "t", "c", "pf", "fl", "no" means vermal, intermediate (paravermis), hemisphere, anterior, posterior, translobular (anterior1posterior), central, parafloccular, floccular, nodular, respectively. The numeral (such as "1" in "vp1-2") counts the cluster from the medial to the lateral side in each category ("it1" is absent).
At E15.5, three major rearrangements occurred in the cl cluster. Firstly, the most medial part of the c-l lineage cluster was recognized as a separate daughter cluster because of its strong Corl2 expression (single asterisk in Fig.  9A2, B2, C2, D2). Secondly, the rostrolateral part of the cl lineage cluster migrated further laterally in areas rostral and ventral to the ml lineage cluster and increased in EphA4 expression (arrowheads in Fig. 9A2). The caudolateral part of the c-l lineage cluster retracted medially. Lastly, the lateral part of the c-m lineage cluster, (Fig.  9A2, B2, double asterisk), migrated laterally at the position ventral to the c-l lineage cluster to separate the ventral portion of the c-l lineage cluster into the rostral and caudal parts (Fig. 9C2, D2).
At E16.5, the separation of the c-l cluster that started at E15.5 became clearer. The most lateral part of the c-m lineage cluster further migrated laterally to separate the rostral and caudal parts of the c-l lineage cluster completely (Fig. 9A3, B3, C3, D3, double asterisk). This lateral migration of the c-m lineage cluster confirmed our previous observation with pcdh10 reporter mice (Vibulyaseck et al., 2017). The rostral part of the c-l lineage cluster was  -4 (va2, va3, va4) . .. 111 .. 11 .. 11 fl5 ";" indicates the gradient in the molecular expression intensity. ".." indicates the change in the molecular expression intensity spread widely in the mediolateral direction (arrowheads in Fig. 9A3, B3) and subdivided into three daughter subclusters that were recognized by different molecular expression profiles (Fig. 9A3, B3). The lateral daughter subcluster migrated laterally and also elongated caudally (blue in Fig 9B3).

The lineage of the c-l cluster at adult
Our previous study (Fujita et al., 2012) has suggested that the E17.5 clusters that were identified as the c-l lineage clusters in the present study (va2-4, ia3-5, ia1-2, it3, ha4, and ip1-2 clusters in the E17.5 cerebellum, Fujita et al., 2012) become zebrin-negative and-lightly-positive stripes in paravermal and hemispheric areas in the anterior and posterior lobules. We tried to confirm the location of c-l lineage stripes directly in the adult cerebellar cortex by using G2A::Ai9::AldocV mice which show tamoxifeninduced birthdate-dependent labeling of neurons with tdTomato as well as labeling of zebrin-positive PCs with Venus fluorescent protein (Fig. 10A). We have reported the dependency on tamoxifen injection timing of the general pattern of PC labeling in adult G2A::Ai9::AldocV mice (Zhang et al., 2020). Here, we mapped E10.5-PCsparse cortical areas by the observation of cerebellar sections of G2A::Ai9::AldocV mice that received tamoxifen at E10.5 (n=2 mice) to identify c-l lineage stripes.
In the adult mouse cerebellar cortex, the striped expression pattern of zebrin (aldolase C) is consistent across individuals and has been identified in detail (Fujita et al.,   Sarpong et al., 2018). Accordingly, zebrin stripes are described here by using the common nomenclature (Fujita et al., 2014;Sarpong et al., 2018). Zebrin stripes are mostly designated with a numeral or an alphabet character followed by "1" or "-" (Fig. 1C). Note that a pair of stripes in the anterior and posterior lobules, which are named differently, together belongs to the same cerebellar module, because they receive common branching olivocerebellar axons (Sugihara and Shinoda, 2004): for example, a pair of 41 stripe in anterior lobules and 51 stripe in posterior lobules form the C2 module and thus considered to be the same or linked stripe, which is designated as "41//51".
In lobules I-V, except for stripe 21 and the medial part of stripe 2-, most of the paravermal stripes (the lateral part of zebrin stripes 2-, stripes b1, b-, 31 and 3-) were E10.5-PC-sparse and consequently identified as the fate of c-l lineage clusters (the right circumscribed areas in Fig. 10B1). However, the most lateral part of stripe 3-and the entire stripe 41 contained a high density of E10.5born PCs. Most of stripe 4-was E10.5-PC-sparse (the left circumscribed area in Fig. 10B1). The most lateral part of stripe 4-and entire stripe 51 contained a high density of E10.5-born PCs. Stripe 5-was E10.5-PC-sparse (the left ventral circumscribed area in Fig. 10B1).
In the simple lobule, most of the medial paravermal area had a low-dense distribution of E10.5-born PCs (double asterisks in Fig. 10B1). This area was likely to be derived from c-m lineage clusters, which had similarly a low density of E10.5-born PCs in the embryonic cerebellum (Fig.  7D). This agrees to the speculation of our previous study that this area originates from E17.5 clusters ic1 and ic2 (Fujita et al., 2012) or cluster ic1-2, which originated from the E14.5 c-m cluster in the present result (Table 2). Zebrin positive stripe 41 and medially adjacent zebrin negative area contained a high density of E10.5-born PCs. On the contrary, stripe 4-, except for its lateral part, was E10.5-PC-sparse. Stripe 4-was also E10.5-PCsparse in crus I (the most left circumscribed area in Fig.  10B1). These E10.5-PC-sparse areas were identified as the fate of c-l lineage clusters (circumscribed in Fig. 10B1) occupied a relatively large portion in the paravermal cerebellar cortex in the rostral cerebellum.
Some tdTomato labeling in the molecular layer inside the circumscribed areas indicated the presence of labeled PC dendrite (left magenta labeling in the inset in Fig.  10B1). However, the number of labeled somata that were located within the section (arrowheads in Fig. 10B1-4) was rather small (12 PCs in Panel B1, roughly 1-2% against the denominator of all PC somata, ;700, recognized in the background inside the circumscribed areas), supporting our description "mostly lacking" and "E10.5-PC-sparse". This percentage (1-2%) was approximately the same as the number of labeled PCs inside the E14.5 c-l cluster (preceding section).
In the rest of cerebellar lobules (lobules VI-X, crus I, crus II, paramedian lobule, copula pyramidis, paraflocculus and flocculus), several areas were E10.5-PC-sparse and identified as the fate of the c-l clusters in the paravermis and hemisphere. These areas included the lateral part of stripe 2-in lobules V-VI (circumscribed in Fig. 10B2), the medial part of stripe 5-in crus II (circumscribed in Fig.  10B2-B3), and multiple zebrin-negative and faintly zebrin positive stripes in the copula pyramidis and adjacent paramedian lobule (circumscribed in Fig. 10B2, B4). Distribution of E10.5-born PCs in the medial paravermal areas in crus I, crus II and paramedian lobule, which are likely to belong to c-m lineage areas (Fujita et al., 2012;Vibryaseck, 2017), appeared a little denser (double asterisks in Fig. 10B2-4) than the c-l lineage areas. Other stripes in the paravermal and hemispheric areas had a rather dense distribution of E10.5-born PCs.
The E10.5-PC-sparse zebrin stripes that were considered to belong to the c-l lineage were mapped on the unfolded scheme of the entire cerebellar cortex with the zebrin (aldolase C) striped pattern ( Fig. 10F; Fujita et al., 2014;Sarpong et al., 2018). The c-l lineage areas were located in zebrin-negative and weakly-positive stripes in the lateral vermis and paravermis in lobules I-V and VIII/ copula pyramidis (red and green in Fig. 10F), and in the medial hemisphere in lobules I-V, simple lobule, crus II and paramedian lobule (blue in Fig. 10F). However, no c-l lineage areas were observed in the central part (lobules VI-VII and apex of crus I) of the cerebellum. Based on the correspondence between the zebrin stripes and cerebellar modules (Sugihara and Shinoda, 2004;Sugihara et al., 2009), the mapped stripes corresponded to the B module (red in Fig. 10F), which projects to the lateral vestibular nucleus (Sugihara et al., 2009;Ruigrok et al., 2015), and most of the C1, C3 and D0 modules, which project to the anterior interposed nucleus (green and blue in Fig. 10F, Table 5). The results demonstrated that the separation and migration of the c-l cluster during development forms multiple stripes that mainly correspond to the anterior and posterior parts of the C1/C3 module (Fig. 10C-F), and also the B module and a part of the D0 module.

Discussion
The present study demonstrated the spatial development of the compartmental organization of the mouse embryonic cerebellum between E14.5 and E17.5. The lineages of the nine E14.5 PC clusters which transformed into 37 clusters at E17.5 were tracked by birthdate-tagging. Furthermore, it was shown that one of the E14.5 continued sparse and weak in Pcdh10 expression in A and C. Single asterisks indicate the most medial part of the c-l cluster or the most medial c-l lineage cluster which had higher expression of Corl2 than the rest of the c-l lineage clusters in A-D. Double asterisks indicate the most lateral part of the c-m cluster or the most lateral c-m lineage cluster which intercalated the c-l lineage clusters in A-D. Arrows indicate the direction of cluster migration. Scale bar in C1 (100 mm) applies to C1-C4. Abbreviations, c-l, c-m, d, ml, names of E14.5 clusters; ha4, ia1-2, ia3-5, ia4, ic1-2, ic3, ip1-2, it2, it3, va2-4, names of E17.5 clusters; C, caudal; D, dorsal; L, lateral; M, medial; R, rostral.V, ventral. Figure 10. The fate of c-l lineage clusters in the compartmentalization of the adult cerebellar cortex. A, Confocal high magnification images of the cerebellar cortex of the G2A::Ai9::AldocV mouse in which tamoxifen was given at E11.5, showing tdTomato expression in Purkinje cells. B, Four coronal sections of the left cerebellum at rostral (B1), central (B2) and caudal (B3, B4) levels in the G2A::Ai9::AldocV mouse that received tamoxifen at E10.5 and sacrificed at P42. Superimposed images of magenta (tdTomato representing E10.5-born PCs) and green (Venus representing aldolase C expression) channels are shown in the left, while only green channel (Venus, aldolase C) is shown in the right. E10.5-PC-sparse areas that were considered to belong to the lineage of the c-l cluster were circumscribed by yellow lines. Arrowheads indicated labeled PC somata that were observed within the circumscribed area. Zebrin stripes were indicated in the right panels. Scale bar (500 mm) applies to B1-B4. Inset in panel B1 shows sparse Purkinje cell labeling at the dendrite (left) and at the soma and dendrite (arrowhead) in the c-l lineage area. C, Schematic of intermittent PC generation in the ventricular zone of the left embryonic cerebellum. D, Schematic of the c-l cluster composed of randomly located E11.5-and 12.5-born PCs in the left E14.5 cerebellum. E, Schematic of separation and migration of c-l lineage clusters (blue, green and red) at E17.5. Gray clusters are non-c-l-lineage clusters that separate the green clusters into the anterior and posterior parts and separate the blue cluster into the anterior and posterior parts later. F, Mapping of the E10.5-PC-sparse areas that were considered to belong to the lineage of the c-l cluster mapped the unfolded scheme of the entire cerebellar cortex with the zebrin striped pattern . Projections of the B module and C1/C3/D0 modules are shown schematically. Abbreviations: 11, 1and so on, aldolase C (zebrin) stripes 11, 1-and so on; I-X, lobules I-X; a-c, sublobules a-c (as in IXa-b); AIN, anterior interposed nucleus; B, C1, C3, D: modules B, C1, C3 and D0; C, caudal; Cop, copula pyramidis; Cr I, crus I; Cr II, crus II; D, dorsal; Fl, flocculus; ha4, ia1-2, ia3-5, ic1-2,ic3, ip1-2, it2, it3, va2-4 names of E17.5 PC clusters; L, lateral; LVN, lateral vestibular nucleus; M, medial; M1, primary motor cortex; Par, paramedian lobule; PFl, paraflocculus; R, rostral; Sim, simple lobule; V, ventral. clusters named c-l differentiated into several zebrin stripes that belong to the C1/C3 module in the anterior and posterior lobules. The results supported our hypothesis that the rearrangement of embryonic PC clusters shapes the compartmentalization of the adult cerebellar cortex which may underlie its modular organization and the dual somatotopy.

Cerebellar compartmental organization originate from the differentiation of early PC clusters
The present study revealed a progressive change in the compartmental organization in the embryonic cerebellum. The number of PC clusters recognized by the marker molecule expression profile increased from nine at E14.5 to 37 at E17.5. This increase was accompanied by a significant change in the spatial arrangement of PC clusters between E14.5 and E17.5, as demonstrated in the separation and migration of the c-l lineage clusters in the present study. Since the clustered compartments of PCs in the late embryonic stage (E17.5 in mice) are mostly comparable to the striped compartments of PCs in the adult (Fujita et al., 2012), our clarification of the compartmental differentiation in the period before E17.5 would lead to a better understanding (see below, "Developmental origin of the dual representation of the somatosensorimotor function in the cerebellum") of adult cerebellar compartmentalization. Compartmental organization at stages earlier than E14.5 was beyond the scope of the present study.
Concerning marker molecules used in this study, we speculate that transcription factors FoxP2 and Corl2 may control the expression of compartment-specific molecules, and adhesion molecule Pcdh10 and receptor tyrosine kinase EphA4 may be involved in cluster formation and cellto-cell connection between highly-expressing neurons (Vibulyaseck et al., 2017;Sarpong et al, 2018). However, the functional significance of these marker molecules in the development of PC compartmentalization has not been fully clarified. Nevertheless, differences in their expression levels among PCs were useful to detect trackable PC clusters in the present study. Based on different expression intensities of these marker molecules, we distinguished PC clusters in the entire cerebellum in embryonic dates between E14.5 and E17.5 (Figs. 2-5, Table 2) The PC cluster organization identified in our study conformed to the PC organization that was identified based on gene expression profiling analysis at E14.5 (Wizeman et al., 2019), except for the Nrgn-positive cluster. The Nrgn -positive cluster, which is located widely above the ventricular zone and under other clusters (Wizeman et al., 2019), was not recognized as a cluster in our present study. Nrgn mRNA is expressed in neuro-progenitor and immature neurons in cerebral cortical culture (Nazir et al., 2018). Therefore, there is a possibility that Nrgn-positive cluster of Wizeman et al. (2019) represents newly-born PCs that are joining one of the other clusters. Besides, six of our nine E14.5 PC clusters m, c-m, c-l, d, ml, and l corresponded to clusters Ebf, Ebf/Calb1, Nrgn/Calb1, Ebf/ Dab1-dorsal, Ebf/Dab1-ventral and Foxp1/Dab1 in Fig.  4H of Wizeman et al. (2019), respectively. It would mean that our PC clusters defined by marker molecule expression and PC birthdate and PC clusters defined by gene expression profiling make one-to-one correspondence, except for the Nrgn-positive cluster (above), at the caudal level of the cerebellum. The correspondence of three of our nine E14.5 clusters (dl, rdl and vl) was not known because they were located at the more rostral level than the section shown in Wizeman et al. (2019). As a whole, the high consistency between our E14.5 cluster recognition and PC gene expression profiling support the cluster assembly that we recognized here represents not an arbitrarily defined structure but an essential organization of the E14.5 cerebellum.

Birthdate-specific neuronal labeling to track the lineage of early PC compartments
Tracing the lineage of early PC clusters in the embryonic cerebellum by birthdate-specific labeling of PCs was the essential method in this study. Since PCs of different birthdates are distributed heterogeneously in different PC compartments (Hashimoto and Mikoshiba, 2003;Namba et al., 2011), birthdate-specific labeling of PCs can be a useful experimental tool to track the lineage of PC clusters. Indeed, the original birthday-specific labeling with the ascl1 gene CreER-LoxP system showed that PCs labeled by tamoxifen given at E10.5, E11.5 and E12.5 are distributed differently in the adult cerebellum (Sudarov et al., 2011). Such a birthdate-specific labeling system is technically more accessible and appears more sensitive than previously established birthdate-specific labeling methods (systemic injection of 5-bromo-2'-deoxyuridine, BrdU; systemic injection of tritium-labeled thymidine,  Sugihara and Shinoda, 2004). 2 Cerebellar nucleus subarea that is innervated by PCs in the stripe (Sugihara et al., 2009 cerebellum and 5-in the posterior cerebellum, Fig. 10F) or the C3 module. The separation of stripe 4-//5-into medial and lateral substripes has been reported in our analysis of the PC birthdates in the adult cerebellar cortex (Zhang et al., 2020). It was noticeable that only the medial substripe of 4-//5-originated from the c-l cluster. Consequently, the lateral substripe of 4-//5-is supposed to originate from a different cluster at E14.5. The most lateral part of the c-l cluster formed a small part of zebrin-negative stripe 5-//6or the D0 module (Sugihara and Shinoda, 2004). The D0 module is the somatosensorimotor module akin to the C1/C3 modules and containing the area involved in the eye-blinking reflex (Attwell et al., 2001). According to the present results, most parts of the D0 module including the entire caudal parts of the D0 module in the posterior cerebellum, seem to originate from different clusters. Thus, several questions remain regarding functional domains delineated by embryonic clusters and adult striped organization. The relationship between the early PC compartmentalization and the mossy fiber projection pattern is also to be studied.