Research Article

Spatiotemporal expansion of primary progenitor zones in the developing human cerebellum

See allHide authors and affiliations

Science  25 Oct 2019:
Vol. 366, Issue 6464, pp. 454-460
DOI: 10.1126/science.aax7526

Close-up of human cerebellar development

Early on, cerebellar development shares similarities across humans, nonhuman primates, and even mice. But differences emerge while development progresses, as cellular and molecular analyses by Haldipur et al. now reveal. The rhombic lip persists longer during cerebellar development in humans than in either the mouse or the macaque and generates a pool of neuroprogenitor cells. Similarly, the ventricular zone of the human cerebellum goes a step further than that of the mouse in developing an additional proliferative layer with outer radial glia cells. Transcriptome analysis revealed detailed similarities and differences between progenitor cells of the developing human cerebellum and neocortex.

Science, this issue p. 454


We present histological and molecular analyses of the developing human cerebellum from 30 days after conception to 9 months after birth. Differences in developmental patterns between humans and mice include spatiotemporal expansion of both ventricular and rhombic lip primary progenitor zones to include subventricular zones containing basal progenitors. The human rhombic lip persists longer through cerebellar development than in the mouse and undergoes morphological changes to form a progenitor pool in the posterior lobule, which is not seen in other organisms, not even in the nonhuman primate the macaque. Disruptions in human rhombic lip development are associated with posterior cerebellar vermis hypoplasia and Dandy-Walker malformation. The presence of these species-specific neural progenitor populations refines our insight into human cerebellar developmental disorders.

Human cerebellar birth defects are common and often cause motor and cognitive disabilities, yet most knowledge of cerebellar development comes from mice. The mouse cerebellum shares many features of lamination, circuitry, neuronal morphology, and foliation with humans. However, compared to the mouse cerebellum, the human cerebellum contains 80% of all brain neurons and has a 750-fold larger surface area, increased neuronal numbers, altered neuronal subtype ratios, and increased folial complexity (13).

Human cerebellar development begins by 30 days after conception and is complete by the end of postnatal year 2 (4, 5). Mouse cerebellar development is nearly complete by postnatal day 15 (P15), after just 19 gestational days. Although human development is protracted, this alone is unlikely to explain species differences. Cerebral cortical neurogenesis differs between these species, with humans having large numbers of outer radial glia in an outer subventricular zone that drive human cortical expansion and gyrification (6, 7). Because neuronal numbers in the cerebellum and cerebral cortex scale across evolution (8), we expected divergent cerebellar neurogenesis programs across species.

Studies of human cerebellar development began before publication of photographic plates (9). Limited histological data are available from 10 postconception weeks (PCW) through late gestation (4, 5, 1012). Developmental neuroimaging atlases have been compiled from in utero magnetic resonance imaging studies from gestational weeks (GW) 20 to 24 (i.e., 18 to 22 PCW) but are of limited resolution (13). Major gaps in available human data correspond to essential cerebellar developmental epochs defined in model vertebrates. Here, we analyze human cerebellar development from 30 days after conception to 9 months after birth (Fig. 1 and table S1), define the timing of developmental events, and provide insight into cellular and molecular programs driving human cerebellar development.

Fig. 1 An outline of human cerebellar development.

Hematoxylin and eosin (H&E)–stained midsagittal sections of the developing human cerebellum. The ventricular zone and the rhombic lip, the two main zones of neurogenesis, are marked with red and black arrowheads, respectively. Primary fissures (10) and secondary fissures (20) are noted with arrows. Scale bars: 100 μm.

Spatiotemporal expansion of human cerebellar progenitor zones

We surveyed morphology from human and mouse cerebellar sagittal vermis sections. Humans and mice both have two primary zones of neurogenesis: the ventricular zone (VZ) and the rhombic lip (RL) (Fig. 1 and fig. S1A). The VZ gives rise to all γ-aminobutyric acid–releasing (GABAergic) populations, including Purkinje cells. The RL gives rise to all cerebellar glutamatergic neurons. Cerebellar nuclei neurons are generated first, followed by granule cell progenitors of the external granule layer, which proliferate, differentiate, and migrate to become granule neurons of the internal granule layer. Unipolar brush cells arise last (14). The human cerebellar anlage between Carnegie stage 12 (CS12) and CS23 (30 to 56 days) resembles the mouse cerebellar anlage in size and shape, from embryonic day 10.5 (E10.5) to E17.5 (Fig. 1 and fig. S1A). However, the mouse VZ thins between E10.5 and E15.5 (fig. S1, B to G), and the RL disappears by birth (E19/P0) (fig. S1A). The human embryonic RL is small, but the human VZ thickens through 10 PCW. After 10 PCW, the VZ thins and the RL expands into an elongated tail-like structure. Between 11 and 13 PCW, the elongated RL thickens, continuing to trail from the growing posterior vermis. Between 13 and 14 PCW, the RL incorporates into the posterior lobule where it forms a densely packed pool of cells evident as late as 36 PCW (Fig. 1).

The mouse external granule layer is evident by E12.5, with granule cell progenitor proliferation driving postnatal cerebellar and foliation expansion (15, 16). We detect initial external granule layer (EGL) formation in the human cerebellum at 8 PCW (Fig. 2G), with primary and secondary fissure initiation apparent at 11 and 13 PCW, respectively. Between 17 PCW and birth (~36 PCW) there is an approximately fivefold increase in human cerebellar volume and folial complexity (17, 18). Peak proliferation in the human external granule layer occurs during the period between 26 and 32 PCW (10).

Fig. 2 The human cerebellar VZ is expanded into a SVZ.

Midsagittal sections of the human embryonic cerebellum stained with KI67 (A to G) and SOX2 (D), reveal VZ expansion. (H) β-III Tubulin (TUJ1) and (I) CTIP1 expression suggest that neuronal differentiation takes place in the SVZ beginning around CS19. (G) External granule layer (EGL) first appears at 8 PCW. (J to L) Mitosis and mitotic radial glia are observed in both the VZ and SVZ as evidenced by phosphohistone H3 (PH3) (J) and phospho-vimentin (pVIM) [(K) and (L)] expression. (M) A significant increase in the proportion of cerebellar basal progenitors [(J) and (K), red arrowheads] is seen between the human and mouse cerebellum, and between CS18 and CS23 (Chi-square, df: 132.5, 2; P < 0.0001). (N) DiO labeling of VZ and SVZ progenitors at CS22 shows radial glial fibers traversing the thickness of the cerebellum. Sections were counterstained using 4′,6-diamidino-2-phenylindole (DAPI). The boundaries of the VZ, SVZ, and pia are marked with green and black bars along the right side of the images and with dotted lines within the images, respectively. Scale bars: 100 μm.

The human cerebellum has a SVZ with basal progenitors

At CS12, the human cerebellar VZ resembles the E12.5 mouse VZ which displays a single zone of SOX2+ KI67+ progenitor cells spanning most of the anlage (Fig. 2A and fig. S1, B and F). By CS14, an emerging SVZ is evident (Fig. 2B). By CS18 and CS19, differentiating (TUJ1+CTIP1+) neurons increase cerebellar anlage size (Fig. 2, H and I). Increased differentiation in the outer SVZ diminishes SVZ size between CS21 and CS23 (Fig. 2, E and F). By the end of embryogenesis, at 8 PCW, only a residual VZ remains (Fig. 2G).

The expanded proliferative zone in the embryonic human cerebellum resembles the SVZ in the developing mouse and human cerebral cortex (19). The mouse cerebellar VZ does not have a SVZ. Instead, ventricular radial glial progenitors extend processes across the nascent anlage from the ventricular (apical) to the pial (basal) surface and undergo mitosis only at the ventricle (Fig. 2, M and N, and fig. S1, L to Q). In the human cerebellum, mitotic phosphohistone H3+ (PH3+) progenitors are found within the VZ and SVZ, indicating the presence of basal progenitors, which we call cerebellar basal progenitors [Fig. 2, J and K (red arrowheads), and M]. Progenitors in both zones exhibit long radial processes which span the anlage thickness (Fig. 2N) and express mitotic radial glial marker phospho-vimentin (Fig. 2K). We see a significant expansion of cerebellar basal progenitors throughout the SVZ between CS18 and CS23 (Fig. 2M), a time point coinciding with increased differentiation, suggesting they could function as additional neurogenic progenitors.

The human cerebellar RL is long-lived and compartmentalized

Species differences in progenitor zone development are not restricted to the cerebellar VZ. We also identified differences in RL morphology, finding substructure in humans, including a split into ventricular (RLVZ) and subventricular zones (RLSVZ), as well as internalization (Fig. 3 and fig. S3, A to G). The mouse RL is a proliferative, transient, dorsal stem cell zone, present between E12.5 and E17.5 and composed entirely of KI67+ and SOX2+ stem cells lacking morphological compartmentalization (fig. S2, A to E). Although it is 5 to 8 cell layers thick, progenitor mitosis (PH3+ and phospho-vimentin+) is confined to the single layer of cells lining the ventricle (fig. S2, G to L). In contrast, the human RL and its remnants are seen throughout gestation and display a more complex proliferation profile (figs. S3, A to G, and S5C).

Fig. 3 The human cerebellar rhombic lip expresses classic markers.

VZ-born Purkinje cells expressing CTIP1 (A and C) and calbindin (B) migrate around the RL (red arrowhead). The RL expresses classic markers such as (D) PAX6, (E) WLS, (F) TBR2, and (G) LMX1A. (H) ATOH1 is expressed by cells exiting the RL into the external granule layer (white arrowhead). Sections were counterstained with DAPI [(A) to (F)] and fast green [(G) and (H)]. Scale bars: 100 μm (white) and 500 μm (blue).

The human RL excludes VZ-derived GABAergic neurons, expresses classic RL markers, and is proliferative even when embedded in the posterior-most lobule of the cerebellum (Figs. 3, A to H, and 4B; and fig. S3, A to K). The human embryonic RL, similar to that of the mouse, consists of KI67+ and SOX2+ cells (Fig. 4, B and C, CS18; and fig. S2, A to F). However, after 10 PCW, the human RL splits into a SOX2+ and KI67–rich RLVZ (Fig. 4, B and C, red asterisk) and a KI67-rich, SOX2-sparse RLSVZ (Fig. 4, B and C, yellow asterisk). The two RL progenitor zones are separated by a vasculature bed, discernible by 11 PCW (fig. S5, A and B, arrows). Cells in the RLSVZ apparently migrate into the external granule layer (Fig. 4, B and C; and fig. S3, H and I, white arrowheads).

Fig. 4 The human cerebellar RL is compartmentalized into ventricular and subventricular zones.

(A) Illustration of the cerebellar regions studied in this figure. The RL (red arrowhead) is expanded into ventricular (RLVZ, red asterisk) and subventricular zones (RLSVZ, yellow asterisk). (B) KI67 expression reveals extensive proliferation in the RL. (C) The RL is compartmentalized into a SOX2-rich RLVZ and SOX2-sparse RLSVZ. (D) Phosphohistone-H3 and phospho-vimentin expression indicate the presence of mitotic ventricular and subventricular-basal progenitors. (E) The proportion of basal progenitors increases significantly after the splitting of the RL (Chi-square, df: 137.8, 2; P < 0.0001). (F) DiI and lentiviral labeling of an organotypic slice of the human cerebellum reveal diverse morphologies of RL ventricular and basal progenitors. Scale bars: 100 μm (white) and 50 μm (yellow).

Radial glial mitosis in the early human RL is confined to cells lining the ventricle (Fig. 4, D and E, CS18), much like in the mouse (fig. S2, G to L). However, after the split of the RL, mitotic progenitors with radial glia–like morphology (PH3+, phospho-vimentin+) are detected in both the RLVZ and RLSVZ (Fig. 4, D and E, 13 PCW). DiI and lentiviral labeling reveal diverse morphologies of RL basal progenitors including uni-, bi-, and multipolar cells with radially and tangentially oriented processes (Fig. 4, A and F).

Although the mouse RL lacks structural compartmentalization, it is molecularly compartmentalized. An interior Wntless (WLS+) and LMX1A+ compartment is continuous with the VZ and an exterior compartment links to the external granule layer (ATOH1+). A gradient of PAX6+ expression exists across the mouse RL, with strongest expression in the exterior compartment. The core of the mouse RL is composed of proliferating LMX1A+ progenitors destined to become posterior vermis granule cell progenitors and unipolar brush cells. Early specified and differentiating unipolar brush cells in the core also express TBR2 (20, 21).

In humans, WLS expression is also largely restricted to RLVZ cells, although scattered expression was seen in RLSVZ cells (Fig. 3E and fig. S4, A to C). LMX1A is expressed throughout the embryonic RL (fig. S4D), in both the RLVZ and RLSVZ at later stages (Fig. 3G and fig. S4, E and F). LMX1A is also expressed in RL-derived cerebellar nuclear and unipolar brush cell populations streaming into the cerebellar core, as well as the choroid plexus epithelium (fig. S4E). A sharp boundary between LMX1A and ATOH1 in RL exiting granule cell progenitors define the anterior limit of the RL (Fig. 3, G and H, and fig. S4, D to J). The posterior limit of the human RL is defined by LMX1A+ MKI67 choroid plexus cells (fig. S4, D, G, and H). PAX6 expression is predominant in the RLVZ, although there is also extensive expression in the RLSVZ, with up-regulation in nascent external granule layer cells streaming from the RLSVZ (Fig. 3D and fig. S4, K to M). TBR2 is expressed throughout the RLSVZ with a few scattered, presumably nascent unipolar brush cells, in both the external and internal granule layer (Fig. 3F and fig. S4, N to P).

Human RL progenitors share some similarities with mice

To provide an unbiased analysis of the molecular programs encoded by human RL progenitors, we profiled the human RL transcriptome using bulk sequencing from laser-capture microdissected RLVZ and RLSVZ (data S1 to S4). Principal component analysis indicated age as the first principal component, explaining 56% of the variance (Fig. 5A). Differential expression analysis between RLVZ and RLSVZ identified alterations in 622 genes (log2 fold change > 1.5 and Benjamini-Hochberg adjusted P < 0.05). The 374 genes up-regulated in RLVZ included CRYAB, SOX2, and WLS, and the 248 genes up-regulated in the RLSVZ included EOMES (Fig. 5B and data S2). We evaluated pathway enrichment of the up-regulated genes and found RLVZ genes were enriched in HIPPO and WNT signaling (Fig. 5C and data S3). RLSVZ genes were enriched in axon guidance and synaptic vesicle cycling (data S4). Several known mouse marker genes for RL and early RL derivates showed similar expression profiles in our samples (22) (fig. S3L). We next compared genes with significant differential expression between the RLVZ and RLSVZ to mouse gene sets identified in the RL and RL-related cells in recent single cell analyses of the developing mouse cerebellum (23, 24). Whereas roof plate–like stem cell genes and a subset of mouse RL genes were differentially expressed in the RLVZ, genes expressed in mouse RL, unipolar brush cell, and granule cell progenitors were seen throughout the RL. Unipolar brush cell genes were highly expressed in the RLSVZ, indicating that it is likely a reservoir for nascent human unipolar brush cells, in addition to other glutamatergic lineages (Fig. 5, D to G, and fig. S3M). Our analyses validate our hypothesis that the spatiotemporally expanded structure in the posterior vermis is indeed the RL, and similarities in gene expression patterns exist between the mouse and human RL notwithstanding differences in structure.

Fig. 5 RNA-seq of human RL compartments.

(A) Principal component (PC) analysis indicates that the largest source of variation among the RNA sequencing (RNA-seq) samples was age, accounting for 56% of the variance in the data. Samples microdissected from RLSVZ are blue and those from RLVZ are red. The numbers beside each circle represent sample age (PCW, postconception weeks). (B) Volcano plot illustrating differential expression of genes in RLVZ versus RLSVZ. Red and blue dots represent genes expressed significantly higher in RLVZ or RLSVZ, respectively. (C to G) Heatmaps of gene expression for each human sample. Samples are grouped by RLVZ (red) and RLSVZ (blue), then by ascending age (9 to 22 PCW). Expression of (C) Hippo pathway genes and [(D) to (G)] top genes for RL-related cell clusters identified by single-cell RNA-seq of the mouse cerebellum from E10 to P14 (23, 24).

RL spatiotemporal expansion may be a human-specific feature

The developing human and nonhuman primate cerebral cortices share an expanded SVZ relative to mice (25, 26). Because the brain weight–to–neuron number ratio does not differ significantly among primates, we expected nonhuman primates to share elaborated cerebellar progenitor zones (27, 28). We analyzed midsagittal sections of the developing cerebella of rhesus macaque (Macaca mulatta; 164-day gestation). The macaque RL at E48 resembles the embryonic CS23 RL in humans (fig. S6, A and M). However, as foliation initiates, the macaque RL regresses in a manner similar to the mouse RL and unlike humans (fig. S6, B and C). By mid to late gestation (E78 to E133), there is no evidence of RL expansion or morphological compartmentation (fig. S6, D to L). This suggests that spatiotemporal expansion of the RL may be specific to humans.

The internalized RL generates the posterior vermis

Human cerebellar volume increases fivefold between 22 PCW and birth, and it becomes highly foliated during the third trimester (24 to 40 GW) (17, 18). Granule cell progenitor proliferation peaks during this period, accompanied by increased external granule layer thickness (5, 10). In mice, cerebellar growth and foliation are driven by granule cell progenitor proliferation between P1 and P14, with deficient proliferation causing external granule layer thinning, reduced foliation, and reduced cerebellar volume (15, 16). Disproportionate reduction in posterior vermis volume is a feature of many human cerebellar birth defects, including Dandy-Walker malformation and cerebellar vermis hypoplasia (29). Mouse models have indicated that RL disruption in the form of precocious differentiation or aberrant migration of progenitors is central to posterior vermis hypoplasia (3032). In mice and humans, 10 cardinal lobules are grouped together into anterior, central, and posterior lobes of the vermis defined by the primary and secondary fissure. In humans, all 10 lobules are first identifiable between 14 and 18 PCW. Development progresses from anterior to posterior, with the increase in posterior vermis size and folial complexity relative to the anterior cerebellum beginning only after 17 PCW (Fig. 6A). The RL and its vestiges are detectable in human cerebella until birth. The longevity of the internalized RL thus correlates with growth and foliation of the human posterior vermis, a cerebellar region associated with human cognition (33).

Fig. 6 Internalized RL may be a feature specific to human cerebellar development.

(A) Model for human cerebellar development indicates that growth of the posterior vermis correlates with spatiotemporal RL expansion. (B to G) Analysis of H&E-stained sagittal sections of the human cerebellum from cases diagnosed with Dandy-Walker malformation (DWM) and cerebellar vermis hypoplasia (CVH) indicates that RL is absent in 50% of cases [(B) and (C)] and is severely diminished in the remaining cases [(D) to (G)]. The anterior (a), central (c), and posterior (p) lobes are indicated in gray, blue, and red, respectively. Colored arrowheads correspond to the colored features in (H). (H) Pie chart representing the absence of RL in 50% of tested samples. Scale bars: 100 μm (black), 0.5 mm (blue), and 1 mm (red).

To determine whether RL abnormalities specifically contribute to posterior vermis hypoplasia in Dandy-Walker malformation and cerebellar vermis hypoplasia, we assessed 16 archival samples aged 17 to 30 PCW (table S1). All show delay or failure of posterior vermis growth. Most show only a partially formed posterior-most lobule (31). Although overall decreased external granule layer proliferation may contribute to this phenotype, peak external granule layer proliferation only begins around 26 PCW (26 to 32 PCW), and hypoplasia of the posterior vermis is evident earlier (Fig. 6, B to H, and fig. S7, A to P) (30, 31, 34). Indeed, Dandy-Walker malformation and cerebellar vermis hypoplasia are routinely diagnosed in utero around 17 PCW (35). The RL is absent in half of our cases and delayed (not internalized) or reduced in size and cellularity in the others (Fig. 6H). Among cerebella lacking a RL, ~62% are older than 21 PCW, when the human RL is normally still present. We conclude that whereas the early RL (<13 PCW) contributes granule cell progenitors to the external granule layer which later proliferate to drive expansion of the anterior cerebellum, the older, internalized RL (>14 PCW) generates granule cell progenitors required to fully elaborate the posterior vermis during mid and late gestation.

Preterm brain injury is associated with cerebellar abnormalities (36, 37). We find RLVZ and RLSVZ are separated by a vascular bed beginning at 11 PCW, when the RL itself is elongated and perhaps more vulnerable to insult. Vascular insults causing RL structural damage from 13 to 14 PCW onward likely contribute to these neurodevelopmental abnormalities. Thus, although mouse studies were essential to spotlight a role for the RL in human Dandy-Walker malformation, the underlying pathological mechanisms can likely never be fully modeled in mice that lack complex RL anatomy. Similarly, our discovery of previously undescribed progenitor populations and the longevity of the RL in the human cerebellum suggest that mouse models of the cells of origin for some groups of medulloblastoma, a cerebellar tumor, may also be inadequate (38). Our studies underline the urgency of further comparative cellular and molecular analyses of human and mouse cerebellar development to better define the value and limitations of mouse genetic models of human neurodevelopmental disorders.

Supplementary Materials

Materials and Methods

Figs. S1 to S7

Table S1

References (3945)

Data S1 to S4

References and Notes

Acknowledgments: We thank B. Lopez, N. Moreno, M. Crosier, Y. Cheng, J. Dobor (HDBR), D. O’Day (UW), A. Sjoboen, J. Millman, D. Dang, D. Dubocanin (SCRI), B. Wicinski (Hof laboratory), A. Thrasher (UCL), A. Duque and L. Selemon (MacBrainResource), and NIHR GOSH Biomedical Research Centre for providing resources and technical help. We thank D. Price (Edinburgh) for recommending HDBR, and S. Tole (TIFR, India) for feedback on the manuscript. Funding: This work was supported by NIH-R01-NS080390 and R01-NS095733 to K.J.M. and R01-N5050375 to W.B.D. P.H. was awarded EMBO fellowship ATSF-431-2016, Burroughs-Wellcome Fund 1018771, Company of Biologists Fellowship DEVTF190393, and a National Ataxia Foundation Young Investigator Research Grant. P.A. was awarded the Newlife Charity for Disabled Children Start-Up Grant SG/17-18/05. Author contributions: Conceptualization: P.H., K.J.M.; Methodology: P.H., K.A.A., P.A., K.J.M.; Software: K.A.A., A.E.T., P.A.; Validation: P.H., L.M.O.; Formal analysis: P.H., P.A., K.A.A., K.J.M.; Investigation: P.H., M.D., C.W., L.M.O., P.A.; Resources: S.B., S.N.L., I.A.G., D.G., S.J.L., D.K., L.M., R.R., H.A.-B., F.R., E.S.F.G.; Data Curation: P.H., K.A.A., A.E.T.; Writing – original draft preparation: P.H., P.A., K.A.A., K.J.M.; Writing – review and editing: P.H., K.A.A., P.R.H., P.A., K.J.M.; Visualization: P.H., K.A.A., P.A., W.B.D., K.J.M.; Supervision: P.H., K.J.M.; Project administration: P.H., K.J.M.; Funding acquisition: K.J.M. Competing interests: The authors declare no competing interests. Data and materials availability: Human material was provided by the Joint MRC/Wellcome (MR/R006237/1) Human Developmental Biology Resource ( and the Birth Defects Research Laboratory (NIH-R24-HD000836 to I.A.G.) and covered by a material transfer agreement between SCRI and HDBR/BDRL, but samples may be requested directly from the HDBR/BDRL. Macaque images were provided by MacBrainResource (; NIMH-R01-MH113257 to A.D. and L.S.). Most sequence data was deposited into dbGaP, the Database of Genotypes and Phenotypes under accession number phs001901.v1.p1, with remaining data available upon request with data use agreement.

Correction (6 December 2019): The surnames of the authors Homa Adle-Biassette and Fabien Guimiot were accidentally misspelled, and the incorrect affiliation was given for author Andrew E. Timms. These errors have been corrected, and affiliations 4 through 13 have been renumbered as 5 through 14 to accommodate the addition of the correct affiliation for Andrew E. Timms.

View Abstract

Stay Connected to Science

Navigate This Article