An expandable embryonic stem cell-derived Purkinje neuron progenitor population that exhibits in vivo maturation in the adult mouse cerebellum

The directed differentiation of patient-derived induced pluripotent stem cells into cell-type specific neurons has inspired the development of therapeutic discovery for neurodegenerative diseases. Many forms of ataxia result from degeneration of cerebellar Purkinje cells, but thus far it has not been possible to efficiently generate Purkinje neuron (PN) progenitors from human or mouse pluripotent stem cells, let alone to develop a methodology for in vivo transplantation in the adult cerebellum. Here, we present a protocol to obtain an expandable population of cerebellar neuron progenitors from mouse embryonic stem cells. Our protocol is characterized by applying factors that promote proliferation of cerebellar progenitors. Cerebellar progenitors isolated in culture from cell aggregates contained a stable subpopulation of PN progenitors that could be expanded for up to 6 passages. When transplanted into the adult cerebellum of either wild-type mice or a strain lacking Purkinje cells (L7cre-ERCC1 knockout), GFP-labeled progenitors differentiated in vivo to establish a population of calbindin-positive cells in the molecular layer with dendritic trees typical of mature PNs. We conclude that this protocol may be useful for the generation and maturation of PNs, highlighting the potential for development of a regenerative medicine approach to the treatment of cerebellar neurodegenerative diseases.

grafted cerebellar progenitors to properly integrate into the recipient circuitry diminishes as the development of the host advances 17 .
Over the past decade, the development of differentiation protocols from pluripotent stem cells has led to the advancement of in vitro generation of neurons 18 , including those of the cerebellum [19][20][21][22] . Potentially, these technical advances might be useful for further developing treatments for degenerative forms of ataxia as they permit use of genetically homologous patient-derived cells, avoiding the rejection issue 23 . Earlier work has shown that functional PNs can be derived from human ES cells, and that these exhibit substantial self-organizing potential for generating a polarized structure reminiscent of the early human cerebellum at the first trimester 19,22 . In addition, PN progenitors from mouse ES cells migrate to the Purkinje cell plate with their axons approaching the cerebellar nuclei in hosts up to E16 20 . But successful maturation and integration of ES cell-derived cerebellar progenitors has not been reported in adult recipients, which present a more challenging environment for grafted cells 17 . Moreover, until now standardization of differentiation protocols of neural progenitor cells (NPCs) has not led to consistent and robust generation of cerebellar neurons from transgenic mouse models and/or human patients with cerebellar disorders. To date, it has remained unclear what is the best strategy to consistently mature PNs derived from pluripotent stem cells at high numbers in vitro, allowing subsequent functional implantation and integration in adult animals in vivo.
Here, we generated PN progenitors from mouse ES cells implementing a cerebellar differentiation protocol of embryoid bodies (EBs) using insulin, FGF2 and cyclopamine so as to fine-regulate the balance between cerebellar neurons and glia cells, while evaluating the expression of cell-specific factors like Engrailed 2 (En2) and Neph3 24 . We subsequently promoted their proliferation and expansion by using an NPC medium containing FGF2, B27 and N2 on a substrate of the neuroprotectant laminin 25 . In addition, we facilitated their maturation in vitro in NS21 medium, which has been shown to optimize the in vitro micro-environment of primary neurons 26 . The maturation potential of these NPCs was tested in vivo in mice with or without host PNs 27 , using a prematurely aging mouse model characterized by neuronal degeneration, inflammation and behavioural disorders. We show that our protocol allows for the generation of an expandable PN progenitor population that can be matured both in vitro and in vivo, paving the way for future studies that focus on the integration of ES cell derived PNs into the intact adult mouse cerebellum.

Results
We adapted existing stem cell-based cerebellar differentiation protocols 20,28,29 so as to obtain high numbers of cerebellar progenitors that can be implanted in vivo in adult animals. We chose to isolate cerebellar progenitors from EBs, because (i) the use of a cerebellar progenitor population allows for the generation of an intermediate and stable cell state 30 and (ii) the number of PNs that can be generated directly from ES cell cultures is limited [19][20][21][22] . To this end, we: 1) maintained and expanded mouse stem cells in ES medium (referred to as stem cell stage); 2) differentiated mouse ES cells as EBs into the cerebellar lineage (referred to as differentiation stage); 3) expanded NPCs for up to 8 passages (referred to as expansion stage); and subsequently, either 4a) induced further neurogenesis of a cerebellar progenitor population in vitro (referred to as in vitro maturation stage), or 4b) implemented integration of an expandable PN progenitor population in vivo (referred to as in vivo maturation stage) (for overview see Fig. 1).

Stem cell stage: Generation and maintenance of ES cells. Mouse ES cells were collected and isolated
as previously described 31 . Mouse ES cells were cultured and maintained 28 , while special attention was paid to obtain sharply delineated ES colonies and an appropriate ES cell density (Fig. 1).

Differentiation stage: Cerebellar differentiation of EBs. Midbrain hindbrain boundary (MHB)
patterning was induced by adding insulin, FGF2 and cyclopamine to ES cells in serum-free floating culture of embryoid body-like aggregates with quick reaggregation 20 (SFEBq) (Supplementary Figure 1). After 8 days in vitro (DIV) this induction was confirmed with immunodetection of the MHB marker En2 in EBs, confirming the cerebellar fate of these cells 32,33 . Since cerebellar differentiation may be sensitive to the size of EBs 20, 21 , we compared En2 expression under various conditions in terms of ES cell density (3 × 10 3 or 3 × 10 4 cells/well or 4 × 10 6 cells/dish) and static or dynamic culture, which are all known to affect the size of EBs 34 . We used the expression of En2, Neph3 and E-cadherin, the latter two of which as markers for cerebellar ventricular zone progenitors 24 , so as to be able to determine the specific culture conditions that led to cerebellar patterning in EBs preparations. Under static conditions, En2, Neph3 and E-cadherin could be readily and homogenously induced in multiple cells contained within EB preparations at a starting population of 3 × 10 3 cells/well (Supplementary Figure 1; Fig. 2a). Under dynamic culture conditions or at higher cell densities, the induction of the expression of these proteins was far less prominent and often limited to clumps of cells randomly distributed within EBs (Supplementary Figure 1). We also used a GFP-Actin 35 ES cell line for fluorescence activated cell sorting (FACS) so as to closely reproduce the results of the GAD67-GFP ES cell line that has been used in the past for cerebellar mouse differentiation studies 20 . When the neural progenitors of the neuro-epithelial rosettes of EBs (with 3,000 cells per well under static conditions) were sorted for Neph3 by FACS, 26.1% of the cells were Neph3-positive (Fig. 2b). Moreover, when we tested the percentage of Neph3-positive cells under static conditions (0 RPM) without cyclopamine for control, we found that none of the cells was Neph3-positive ( Fig. 2c), highlighting the impact of cyclopamine-dependent Sonic hedgehog (Shh) inhibition on differentiation of ES cells into MHB tissue 36 . In contrast, the progenitors that were Neph3-positive comprised only 3.1% of the total under dynamic conditions at 100 RPM with 3,000 cells per well with cyclopamine (Fig. 2d). It should be noted that the 26.1% of the Neph3-positive actin-GFP cells were measured at 13 DIV and that this percentage decreased over time. Indeed, when quantified by FACS in wild type cells, the percentage of Neph3-positive progenitor cells gradually decreased from 59.6% at 11 DIV, to 41.0% at 13 DIV, to 12.6% at 15 DIV and to 3.4% at 17 DIV (Fig. 2e). We tested Neph3 expression in both wild-type cells and actin-GFP cells, because the FACS data obtained by we and others 20 indicate that it is technically easier to discriminate between Neph3-positive and Neph3-negative labeling with wild-type cells than with GFP-expressing cells (for wild-type cells see Fig. 2c and e; for actin-GFP cells see   Expansion stage: proliferation of NPCs. After having the SFEBq culture for 11 to 15 days in differentiation medium, we transferred the EBs to NPC medium 29 on 10 cm 2 6-well plates with laminin-coating so as to promote attachment of the EBs (Fig. 3). In this proliferative stage, the cell population could be expanded and passaged up to eight times (P8) with preservation of NPCs, indicated by positive labeling for progenitor markers Nestin and β-tubulin III 37,38 (Fig. 3a). The β-III-tubulin-positive population significantly decreased after 7 passages (P = 2.17 × 10 −8 , from P2 to P6: 29.96 ± 1.39% of total cells as normalized to DAPI), while that for Nestin-positive significantly decreased after 8 passages (P = 1.26 × 10 −6 , from P2 to P7: 90.25 ± 0.89%) (Fig. 3b). Consequently, passaging and preservation of NPCs was performed for up to 6 passages.
Another approach involved directly sorting Neph3-positive cells with FACS from EBs and subsequently transferring them to wells. However, this did not result in stable, viable cultures of Neph3-positive progenitors (data not shown). Likewise, EBs cultured in NPC medium on plates without laminin-coating or EBs cultured in neuron-specific media, such as NS21 or DMEM/F12/FCS/N2/D-Glucose 20 , did not result in sufficient attachment of the EBs, and as a consequence did not yield viable cultures. Finally, when we tried to culture EBs that were subjected to a differentiation stage shorter than 11 DIV or longer than 15 DIV, the attachment to the laminin-coated 6-well plate surface was also insufficient to achieve prominent attachment or expansion.
Maturation in vitro stage: maintenance of functional neurons in NS21 medium. For the maturation stage, we generally used NPCs that were expanded for 2 to 6 passages. These cells were seeded on polyornithine-laminin coated coverslips and cultured as progenitor neurons and glia cells in NS21 medium with AraC 26 . This procedure yielded a considerable number of neurons and it confirmed the standardization of the cell state. The cells expressed calbindin, NeuN or glial fibrillary acidic protein (GFAP), which indicate PNs, cerebellar interneurons and Bergmann glia cells, respectively [38][39][40] (Fig. 4a,b). The calbindin-positive cells, which displayed arborized dendrites with spines typical of PNs, also expressed other late PN markers, such as those for parvalbumin and inositol 1,4,5-trisphosphate receptor (IP3R) (Fig. 4c,d). When NPCs were cultured in this manner, all cell types were randomly distributed across the coverslip ( Fig. 4a-d). If the cells were cultured in DMEM/FCS/ N2/D-Glucose medium 20 rather than NS21 medium, the ratio of calbindin-positive to GFAP-positive cells was significantly lower (P = 9 × 10 −4 ) (Supplementary Figure 2). Likewise, when we used our NS21 medium without the addition of AraC, which is known to prevent the astrocytes from overgrowing the neuronal population 41 , we found that astrocytes became more prominent while neurons became scarce in culture.
The ratio of the various cell types did not significantly depend on the day of the differentiation stage at which the EBs were collected. More specifically, the ratios of calbindin-positive to GFAP-positive cells (Fig. 4e), GFAP-positive cells to DAPI-positive cells (Fig. 4f), calbindin-positive cells to DAPI-positive cells (Fig. 4g), and NeuN-positive to DAPI-positive cells (Fig. 4h) were relatively stable throughout the period in the differentiation stage during which the EBs were collected (days 11 to 15). These data suggest that it is the specific type of medium used at the maturation stage, rather than the temporal course of events during the differentiation stage, that determines the ratio of cell types.
To assess the level of maturation of the NPC-derived PNs after 15 days of culturing on coverslips, we compared the extent of their dendritic arborisation with that of PNs of dissociated cerebellar cells obtained from E16-18 mouse cerebella (Supplementary Figure 3a-c). The calbindin-positive PNs derived directly from the NPCs (NPCPN) showed significantly fewer secondary dendrites (P = 2 × 10 −4 ) and significantly smaller cell bodies (P = 9.9 × 10 −5 ) compared to those of calbindin-positive PNs of dissociated cells (PN), whereas the number of primary dendrites was not significantly different (P = 9.3 × 10 −2 ), (Supplementary Figure 3d-f). In addition, the PNs had a significantly higher number of secondary dendrites (P = 2 × 10 −4 ) than in dissociated PNs in co-culture (CPNs) and also a significantly larger size of the cell bodies (P = 9.9 × 10 −5 ) than in ES-cell-derived NPCs in co-culture (GFPPN). Together, these data suggest that molecular factors produced by NPCs and/or direct contacts between cells can negatively affect the dendritic arborisation and soma growth of dissociated PNs.
Interestingly, upon direct transfer of EBs to coverslips coated with poly-ornithinine/laminin in NS21 medium -that is without passaging through the expansion stage -we could also identify a population of calbindin-positive cells that showed dendritic arbors (Supplementary Figure 4). The dendrites of these cells did not show spines, but the number of primary and secondary dendrites was 3.7 ± 1.1 and 8.3 ± 2.4, respectively, which was comparable to those of NPC-derived PNs. These data indicate that PNs directly obtained from EBs have a similar dendritic arborisation potential as passaged NPCs.
Subsequently, we characterized the electrophysiological properties of PNs with in vitro patch-clamp recordings to determine their viability (Fig. 5a). After a maturation of 13 to 18 days in NS21 medium, NPC-derived PNs showed spontaneous intrinsic simple spike activity of 3.73 ± 1.24 Hz (n = 21 cells) and upon membrane depolarization with current injections, simple spike activity could reach a frequency of 11.1 ± 2.5 Hz. On average the mature PNs on coverslips showed an input resistance of 402 ± 30 MΩ, a resting membrane potential of −50.9 ± 2.7 mV, and a capacitance of 57 ± 6 pF. Likewise, when we cultured NPCs in 3D by depositing cells suspended in BD-Matrigel (1:2 dilution ratio) 42 into transparent µ-wells 43 , NPC-derived PNs also showed intrinsic simple spike activity, albeit at a lower frequency (0.4 ± 0.23 Hz, n = 8 cells) (Fig. 5b). The PNs in the µ-wells showed an input resistance of 351 ± 25 MΩ, a resting membrane potential of −67 ± 3.2 mV, and a capacitance of 71.2 ± 7.8 pF. The NPC-derived PNs in the µ-wells were calbindin-positive, but not NeuN-positive, confirming the Purkinje cell identity at 10 to 30 DIV (Fig. 5c). In contrast to the PN preparation on coverslips, the 3D culture configuration allowed us to perform intracellular biocytin labeling of patched neurons with post-recording immunocytochemical confirmation of calbindin marker expression (Fig. 5d). . Note that the percentage of NPCs drops after passage 6. Error bars represent SEM. Asterisks indicate significance levels comparing the average of a passage between all passages via one-way ANOVA followed by Tukey's with statistical significance set at P ≤ 0.01: *P < 0.01, **P < 0.001, ***P < 0.0001.  for the integration of progenitor PNs are still active, we first evaluated whether NPC-derived PNs would be able to migrate and mature in the recipient cerebellum of newborn mice 45 . For these transplantations, we prepared cell suspensions of NPCs (5 × 10 4 cells/µl as measured with a Biorad ® automated cell counter) in NS21 medium after 2-6 passages at the expansion stage and injected them directly into the brain without any maturation in vitro (i.e. without the in vitro maturation stage outlined above). Eight wild-type mice at postnatal day 1 received three injections of approximately 1 µl of NPCs (i.e. 5 × 10 4 cells/µl) per injection at random depths in the vermis and/or hemispheres of their cerebellum. Thirty days after grafting, six out of the eight mice (75%) showed GFP labelling in their cerebellar cortex (Supplementary Figure 5). Only 0.17 ± 0.06% of the injected NPCs were retrieved, confirming that only a small fraction of the transplanted cells exhibit long-term survival in the postnatal cerebellum. Of the NPC-derived cells that survived grafting, 9.12 ± 1.66% showed the typical morphological presentation of maturing PNs. These cells expressed late PN markers including calbindin and parvalbumin ( Supplementary  Figure 5a and b). Moreover, the dendritic trees of these PNs were extensively arborized and contained spines, which received VGLUT2-positive inputs (Supplementary Figure 5c), indicating the presence of inputs from climbing fibers. Importantly, GFP-positive-calbindin-positive axon terminals were distinguishable along the soma of SMI32-positive glutamatergic neurons in the cerebellar nuclei (Supplementary Figure 5d). These data suggest that NPCs mature and can in principle form normal connections in the developing cerebellum 17,20 .
Next, to find out whether PNs can also survive and mature in an adult host, actin-GFP-positive NPCs in cell suspension (1.5 µl with 5 × 10 4 cells/µl) were injected into the cerebellum of 16 adult wild-type mice aged 4-10 months old ( Fig. 6 and Supplementary Figure 6). Thirty days after grafting, 11 out of the 16 mice (68.75%) showed GFP labelling in their cerebellar cortex (Fig. 6a). The average percentage of surviving GFP-positive cells was 2.8 ± 0.47%, which was slightly higher (P = 0.016) than that found in the early postnatal animals. In all 11 mice with surviving grafted cells, we found morphologically matured PNs, as recognized by the expression of calbindin and parvalbumin, the presence of a developed dendritic tree with spines in the molecular layer, and the morphology of axons projecting towards the white matter (Fig. 6b). Yet, some axons of the grafted PNs showed varicosities forming a terminal club (Fig. 6c), which may reflect either the failure of the axon to traverse through the white matter 46 or a sprouting process that occurs at an infraganglionic plexus 47 .
The  Figure 7). The percentage of calbindin-positive actin-GFP-positive PNs was slightly lower (P = 0.027) than that in the early postnatal animals (9.12 ± 1.66%). Whereas some of the GFP-positive cells prominently expressed calbindin in their dendritic trees (Supplementary Figure 8), others showed no or only very weak calbindin expression, despite displaying a PN-like morphology (Fig. 6b). The majority of the GFP-positive grafted PNs (92.57%) were located in the molecular layer, but some could also be detected in the Purkinje cell layer (1.01%), granular layer (6.08%), or even white matter (0.34%), presumably depending on the migratory route they took following injection into the cerebellum 48 .
Some of the grafted cells could also be situated outside the cerebellar cortical parenchyma. However, we did not observe grafted PNs or NeuN-positive neurons in the cerebellar nuclei, despite the fact that some of the injections were aimed at that locus (Supplementary Figure 6). Actin-GFP-positive cells injected into this area most probably died, as revealed by positive labelling for apoptotic marker caspase-3 (Supplementary Figure 9). Interestingly, despite its low number of surviving cells, the cerebellar cortex did not show caspase-3-positive cells, which may point towards processes such as autophagy for removal of part of the grafted progenitors 49 .
To find out to what extent NPCs matured into PNs in vivo due to cell fusion with host PNs nuclei, we injected GFP-NPCs in the cerebella of CDK9-mCherry-knockin mice 50 (Fig. 6e-g). The cell-body of PNs of CDK9-mCherry-knockin mice prominently express a fluorescent form (mCherry) of CDK9, which is involved in transcription by RNA polymerase-II (Fig. 6e,f). None of the GFP-positive-PNs did express CDK9-mCherry (Fig. 6g), indicating that the NPCs maturation into PNs in vivo described above did not occur due to cell fusion with host PNs nuclei.
To evaluate the effect of age of the host cerebellum on NPC maturation, the percentage of calbindin-positive cells of the total actin-GFP-positive cells were compared among 4-month-old and 10-month-old mice, which included both the normal hosts and the CDK9-mCherry-knockin mice (Fig. 6i). Although the older mice (i.e. 10-month old) showed a downward trend in NPC expression of calbindin with respect to that in the younger mice (i.e. 4-month old), the difference (between 2.26 ± 1.07% and 4.81 ± 0.96% of calbindin-positive to GFP-positive cells) was not significant (P = 7 × 10 −2 ).
To find out whether the ES/NPC-derived grafted PNs received synaptic inputs from climbing fibers and parallel fibers in adult hosts, we investigated 30 days after grafting whether GFP-positive PNs received inputs from VGLUT2-positive and/or VGLUT1-positive boutons, respectively 51 . Both appositions were readily observed (Figs 7 and 8), confirming that the formation of synaptic inputs onto grafted PNs is possible 9,16 . Some of the GFP-positive PNs that expressed calbindin and received input from VGLUT2-positive terminals were located in the Purkinje cell layer, but many of them were located in the molecular layer (Fig. 7a-e). However, even when grafted cells were not located in the Purkinje cell layer or not properly oriented in the molecular layer, GFP-positive PNs received VGLUT2-positive inputs at both the proximal and distal parts of their dendrites (Fig. 8a-d). Likewise, VGLUT1-positive boutons were also identified on both the proximal and distal dendrites of GFP-positive PNs (Fig. 8e,f). Interestingly, some of the VGLUT2-positive and VGLUT1-positive terminals were also apposed to the cell-bodies of GFP-negative host PNs located in the molecular layer (Fig. 7d); possibly this subcellular displacement reflects the impact of proteolytic enzymes secreted by the migrating cerebellar progenitors 48 .
Finally, to examine whether NPCs can mature in a mouse cerebellum that suffers from comprehensive resident PN death, we also injected actin-GFP-positive NPCs in the cerebellum of 8-month old L7cre-ERCC1 KO mice, which lack resident PNs and show severe ataxia 27 (Fig. 9). The average percentage of surviving GFP-positive 48 ± 0.41%, which was comparable to wild-type animals (2.8 ± 0.47%). Additionally, 3.4 ± 0.72% of the grafted actin-GFP-positive cells also expressed calbindin, which was also consistent with the results in wild-type animals (3.66 ± 0.79%). These calbindin-positive actin-GFP-positive cells were also predominantly present in the molecular layer (Fig. 9a,b) and the layer that used to contain the host Purkinje cells (Fig. 9c). Our results indicate phenotypic similarity in the maturation of ES-cell-derived NPCs into PNs in wild type and L7cre-ERCC1 KO mice, highlighting that our protocol allows for the maturation of PNs in the cerebellar cortex of both healthy and ataxic mice suffering from neurodegeneration.

Discussion
In this study, we present a protocol to obtain a cerebellar neural progenitor cell population from mouse ES cells, which display maturation in vitro as well as in the healthy and PN deficient adult mouse cerebellum in vivo. The generation of stable populations of cerebellar NPCs was achieved by combining and adapting previously reported protocols for the generation of stem cell cultures 28 , neuronal progenitors 29 , and cerebellar differentiated cells 20 . By testing the cerebellar maturation potential of NPCs both in vitro and in vivo, we were able to show that the cerebellar NPCs generated in our protocol form a promising cell source for in vivo cerebellar transplantation. Our protocol yielded up to 59.6% PN progenitors as quantified by Neph3, which was more effective than most, if not all, other protocols published so far. For example, the amount of PN progenitors from total cultured cells as estimated by the expression of the markers L7, IP3R1-Calbindin, and Neph3 were 1 to 11% 36, 52 , 0.45% 21 , and 5 to 27.9% 19,20,22 , respectively in prior studies. One of the main reasons why our approach is probably more efficient is that we collect EBs from days 11 to 15 during the differentiation stage without severely decreasing the potential to later mature as PNs in vitro or in vivo 53 . While PN progenitors could not be sorted, EB-derived cerebellar NPCs could be expanded and cryopreserved for up to six passages, allowing storage of large numbers of cerebellar NPCs for maturation in vitro and in vivo without having to initiate differentiation from ES cells every time. Still, the PNs obtained with our protocol also had their limitations, particularly in relation to their integration in the adult brain.
The cerebellar differentiation stage was achieved by focusing on specification of the MHB-region 19,20,23 . Mouse ES cells could be induced in SFEBq culture with the addition of insulin, FGF2 and cyclopamine at 0, 1 and 7 DIV, respectively 20 . Screening of engrailed-2-positive cells in SFEBq culture at 8 DIV indicated that the effectiveness of external and internal caudalizing signals not only depended on the signaling pathways themselves, but also on the size of the EBs. As EB size increased beyond 3,000 cells, engrailed-2-positive cell numbers did not increase proportionally. We expected convection to overcome the penetration problem (or diffusion limitation) of rostro-caudal and dorso-ventral patterning signals into EBs, but we found that this strategy did not result in a larger engrailed-2-positive cell population. This was probably due to the sensitivity of differentiating EBs to external cues, such as stress or dynamic culturing, which led to altered nutrient limitation, highlighting the sensitivity of cerebellar progenitors to spatial and temporal signals 54 .
The NPC expansion stage was improved by studying the expression of the PN progenitor marker Neph3, which served as an ideal marker to readily identify Purkinje progenitors in SFEBq culture 19,20,22,24 . We found that Neph3 expression in neural rosettes of EBs depended on the cell line. The highest level of expression of Neph3-positive cells occurred at 11 DIV for EBs of the wild-type line vs. 13 DIV for EBs of the GFP-actin line. Since Neph3-positive cells sorted through FACS were difficult to maintain alive or keep in culture after isolation 22 , we collected cells directly from EBs for NPC culture and expansion while using Neph3 expression as a parameter to determine the optimal period. Once specification of the MHB-region was reached, neuronal and glial cell progenitor populations were born spontaneously and required no further treatment other than isolation by replating on laminin in NPC medium 29 . Laminin, which can act as a key maintenance component of the basement membrane surrounding epithelial cells during early stages of embryonic development 55 , served as an effective surface coating following neural differentiation 25,56,57 . FGF2, which can induce rhombomere-1 activity 58 , was also critical in that it helped NPCs to maintain their neuronal progenitor functionality, probably by keeping the population mitotically active and by preventing the initiation of terminal differentiation 59 . Our results showed that this influence of FGF2 could be sustained in NPC medium for up to 6 passages before the downregulated expression of neural progenitor markers nestin and β-III-tubulin. Thus, the NPC protocol described here yields an accessible and large cerebellar progenitor population, which allows for standardization of cerebellar differentiation 30 .
The in vitro cerebellar maturation of NPCs was accomplished by adjusting the culture micro-environment. Using NS21 medium 26 , we generated a suitable in vitro culture micro-environment for NPCs on coverslips by allowing partial maturation of intrinsically-active cerebellar neurons. The firing frequency (3.73 ± 1.24 Hz) and input resistance (402 ± 30 MΩ) of the cells obtained with our protocol were considerably lower than those of cultured primary PNs from acutely dissociated cerebella (54 ± 5 Hz and 900 ± 75 MΩ, respectively) 60 . Regarding spike activity, the difference may be partly accounted for by the considerably higher amount of synaptic input express CDK9-mCherry, indicating there is no fusion with resident cells. Arrowhead points to CDK9-mCherry expression inside and around nuclei of resident cells (n = 3 mice). (h) Quantification of calbindin-positive (n = 11 mice) and NeuN-positive (n = 16 mice) cells over a total actin-GFP-positive cells. All CDK9-mCherry micrographs were obtained from sagittal sections. (i) Percentage of calbindin-positive cells over a total number of actin-GFP-positive grafted cells plotted against the age of the host mouse cerebellum (n = 6 for four-monthold mice; n = 5 for ten-month-old mice). Chart shows the mean and SEM with statistical comparison assessed via one-way ANOVA followed by Tukey's with statistical significance set at p ≤ 0.01. Scale bars: 500 μm (a); 20 μm (b, c, f and g); and 40 μm (d and e). ML: Molecular layer; PCL: Purkinje cell layer; GCL: Granule cell layer. *all micrographs represent coronal sections unless otherwise stated. from granule cells received by primary PNs compared to NPC-derived PNs. Regarding the lower input resistance, the difference may be due to differences in the recording temperatures or the use of cesium in the recording solution 61 . We also investigated the functionality of the cells through electrophysiological means in an adapted 3D culture system, in which the NPCs were embedded in Matrigel 42 and deposited in transparent µ-wells 43 . This combination allowed for spontaneous development of NPCs, including aggregation of calbindin-positive cells with spontaneous spike activity. The lower spiking frequency and membrane properties of PNs in the µ-wells as compared to those on the coverslips might be related to a different maturation level of the PNs cultured in the µ-wells [62][63][64] . The Matrigel-µ-well culture model proved especially useful for morphological reconstruction of the patched neurons, as it allowed intracellular filling of the PNs with biocytin. This was not possible with the coverslip approach because the neurons were more prone to bursting when biocytin was injected in the neurons on the coverslips than when biocytin was injected in the neurons embedded in Matrigel. Future cerebellar disease in vitro modeling should actively include 3D cell culture systems, which are increasingly important for culture of cells from patients 19,65,66 .
The maturation of NPCs in vivo was established via transplantation into the cerebellum of P1 or adult mice. Previous studies have shown the successful engraftment of ES-derived cells exclusively in embryonic and late postnatal cerebella 20,52 . Here, we show maturation of ES-derived NPCs injected in not only early postnatal, but also adult, cerebella. Indeed ES-derived PNs could mature in the adult cerebellum of both wild-type mice and L7-ERCC1 mutant mice that lack host Purkinje cells 27 . ES-derived PN progenitors and PN progenitors from the embryonic cerebellum show the correct integration when transplanted into the developing cerebellum 9,17,67 . Previous studies have reported that host and grafted Purkinje cells may compete to occupy the host molecular layer 4,7,15 , which could in principle lead to higher maturation of grafted PNs in a cerebellar environment depleted of PNs. However, our experiments did not show a better maturation in the L7-ERCC1 mice, which suggests that other factors, such as those related to the intrinsic constitution of the grafted cells, also contribute.
The integration of transplanted NPCs obtained with the current protocol requires further work, because there was a low survival rate of injected cells while the vast majority of the grafted PNs occupied ectopic locations in the host molecular layer and showed no signs of cortico-nuclear projections. One of the significant challenges may be to circumvent endogenous cues, such as Reelin, in the adult cerebellar environment that are known to prevent the correct migration of cerebellar progenitors 17,68,69 . Indeed, when this challenge is not met, as in our current study, surviving grafted PNs will often remain displaced within the upper two-thirds of the host molecular layer and most of them will have inverted dendritic trees and local axonal terminal arborizations with varicosities 4, 5, 9, 15, 48 . In contrast to grafted cerebellar progenitors in developing cerebella, grafted cerebellar progenitors in the adult cerebellum do not migrate in high numbers to the Purkinje cell layer 17 and do not provide prominent projections to the cerebellar nuclei 70,71 , as also was the case in the current study. Possibly, both host and grafted Purkinje cells are particularly sensitive to toxic agents or critical environmental conditions 72 , because of their high intrinsic activity and metabolic demands 73 . Yet, when embryonic Purkinje cells are transplanted into an adult environment, they have in principle the ability to adopt several strategies to induce modifications in the host environment in order to create a permissive environment and make their way to the target position 48,74 . Grafted embryonic PNs can induce modification in gene expression of the host glia in order to facilitate their radial migration in the adult molecular layer, which is called "adaptive rejuvenation" 74 . Similarly, migrating neurons can induce secretion of proteolytic enzymes so as to disrupt components of the host extracellular matrix, such as laminin of the basal lamina, and enter the host molecular layer from the interlobular space 48 . As a consequence, disruption of the recipient texture might even lead to displacement or removal of the host Purkinje cells. Moreover, because of this variety of interactive mechanisms, heterogeneous populations of cerebellar progenitors might be more efficient in inducing a successful transplantation than homogeneous subpopulations of Purkinje progenitors, as it increases the likelihood of containing the proper facilitating ingredients 20 . As mentioned above, isolated PN progenitors in this study were not viable after sorting. The heterogeneous nature of the NPC population in our study may have had several consequences. For example, it may have facilitated the co-occurrence of calbindin-positive and calbindin-negative, GFP-labeled PCs following grafting. Our NPCs were probably subject to differential heterogeneous developmental events as opposed to PC progenitors derived from embryos, which traverse a homogeneous development with a well-synchronized expression of cellular markers. Conversely, the heterogeneity of the NPC population might also have protected some of the implanted PNs and facilitated the activation of endogenous compensatory functions in the recipient cerebellum 67 . This possibility is also in line with the fact that we found no evidence for fusion of grafted and host cells. Indeed, triple positive GFP-calbindin-mCherry cells were not observed following grafting in the cerebella of mCherry-knockin mice, nor were healthy resident Purkinje cells observed following grafting in the cerebella of L7cre-ERCC1 knockout mice. Moreover, the morphology of the dendritic trees of differentiated GFP-positive-calbindin-positive PNs as observed 24 days after grafting was relatively immature as compared to host controls, highlighting the divergent developmental route. Likewise, different from embryonic cerebellar transplants performed with progenitors that have been exposed to the presence of afferences 48 , ES-derived NPCs are not exposed to an endogenous environment until grafting. As a consequence, a maturation period of 24 to 30 days may also be too short for ES-derived PCs to acquire the full extent of segregation of afferent inputs, including not only those of interneurons but also of climbing fibers. Thus, studies with NPCs maturing for longer periods of time are probably needed to evaluate both the development and maturation of ES derived NPCs and the formation of stable connections in the mouse cerebellum. Future studies will be required to resolve the extent to which proteolytic enzymes are responsible for displaced resident Purkinje cells 48,74 and/or injection-induced local hypoxia 75 and to what extent progenitors and host cerebellum can be manipulated to enhance efficiency of integration. The occurrence of displaced or malformed Purkinje cells by itself may not necessarily have unavoidable negative consequences for possible therapeutic use. If the transplantation of ES-derived PC progenitors could replace missing PCs in human ataxia and if their synaptic inputs and outputs integrate in the cortical and downstream circuits, the secondary pathology resulting from the grafting process might have only a relatively minor functional impact. Hence, provided we can improve survival, migration and integration of NPCs in the host cerebellum, and limit the adverse impact of the grafting procedure, ES-derived NPCs might be considered as a source for human therapy in the future.

Materials and Methods
Animals. All experimental procedures involving animals were approved a priori by an independent animal ethical committee (DEC consult, Soest, The Netherlands) and/or by the national authority (Centrale Commissie Dierproeven, The Hague, The Netherlands) and performed in accordance to Dutch legislation and institutional guidelines (Erasmus MC, Rotterdam, The Netherlands and Netherlands Institute for Neuroscience, Amsterdam, The Netherlands).
Cerebellar differentiation protocols were performed on: i) actin-GFP ES cells derived from embryos obtained from crossings between a female wildtype 129 Sv mouse and a male heterozygous actin-GFP mouse 35 with a C57BL/6 background; and ii) wild type ES cells derived from embryos obtained from crossings between female and male C57BL/6 mice. Injections in the adult cerebella involved wild types obtained from a cross between female and male C57BL/6, Purkinje-neuron-specific L7cre-ERCC1 homozygous knockouts 27 and CDK9-mCherry knockin 50 mice with a C57BL/6 background. ES derivation and culture conditions. The ovarian cycle of female mice was synchronized by intra-peritoneal injections of pregnant mare serum gonadotrophin and human chorionic gonadotropin (both 150 µl, 50 U/ml, both one injection with 48 hrs interval). Following the last injection, each female mouse was mated with a male mouse. Females that were found with a vaginal plug ~20 hrs following the start of mating were euthanized at E3.5 and their embryos were collected. For ES derivation 31 , E3.5 blastocysts of actin-GFP mice and wild type mice were placed into culture dishes coated with gelatin (0.2%) and irradiated mouse embryonic fibroblasts (MEFs) in ES cell medium containing DMEM, 15% fetal calf serum (FCS, Lonza), PenStrep (PS, Sigma-Aldrich), 1 mM non-essential amino acids (NEAA, Sigma-Aldrich), 50 nM β-mercaptoethanol (Sigma-Aldrich), leukaemia inhibitory factor (LIF, Sigma-Aldrich), MEK inhibitor (4 µM, Stem cell technologies) and GSK3 inhibitor (3.3 µM, Stem cell technologies). One week after collecting embryos, the outgrowth of the inner cell mass was enzymatically split and plated by returning them to the same, but refreshed, culture conditions. After 3 passages, inhibitors for MEK and GSK3 were removed from the culture medium.
Neuronal differentiation culture. Unless stated otherwise, FGF2 (20 ng/ml, Life Technologies) and Cyclopamine (10 µM, Bio-connect, The Netherlands) were added to SFEBq cultures on days 1 and 7, respectively 20 . SFEBq and petri dish (Greiner Bio-One) cultures were placed in an incubator (Heraeus, Germany) for up to 17 days under static conditions and/or dynamic conditions (100 RPM, IKA Vibrax VXR, Germany), as indicated.
For 3D culture, µ-wells (3 µl per µ-well, Vers3D TM , Screvo B.V., The Netherlands) 43 made of Epoxy (Epotek stycast 1264 A/B, Epoxy Technology, Belgium) were seeded with 3 µl per µ-well of a cell suspension consisting of NPCs in NS21 diluted in Matrigel (1:2, BD Biosciences) 42 . The final cell concentration in the mixture was ≥1 × 10 7 cells/ml. The cell/Matrigel mixtures in the µ-wells were incubated at 37 °C for 5 min before 1 ml of pre-warmed NS21 was added to completely cover each µ-well in 24-well plates.
Flow cytometric analyses of cell aggregates. Cells were sorted using a BD FACSARIA III (BD Biosciences, San Jose, CA). Single-cell suspensions of EBs in 96-well plates were obtained by collecting all EBs of one 96-well plate in one 15 ml tube, centrifugation at 1500 RPM for 2 min, followed by removal of 14 ml of supernatant. Then, 0.5 ml of Accumax (Chemicon, Germany) was added to the EBs in 0.5 ml of medium and the contents were mixed with a 1 ml pipette and incubated for 15 min at room temperature. For Neph3-positive cells isolation and measurement, EBs were dissociated with Accumax as described above on 11, 13, 15 and 17 DIV and filtered through a cell strainer (BD Biosciences). Isolated cells were incubated in Neph3 monoclonal antibody (30 min, Sheep 1:200, R&D Systems) and labeled with a PE-conjugated secondary antibody (30 min, BD Biosciences). Doublets were excluded from the analysis. Sorted cells were gathered in ice-cold DMEM/F12/ N2/10%FCS or NPCs medium and aggregated again in low-cell adhesion 96-well culture plates or 24-well plates.
Electrophysiology. Cells cultured on coverslips were recorded at room temperature (21 ± 2 °C), during which the extracellular bath solution contained (in mM): 124 NaCl, 2.5 KCl, 1.25 Na 2 HPO 4 , 2 MgSO 4 , 2 CaCl 2 , 26 NaHCO 3 and 20 D-Glucose (osmolarity: 295 Osm/L). Because NS21 medium has a lower osmolarity (255 Osm/L) than the extracellular bath solution, cultured neurons were gradually brought to equilibrium with the solutions before recording. This was done by adding 200 µL of extracellular solution to the cultured neurons in NS21 medium and removing 200 µL of the mixed solution every 3 minutes for a total of 20 min. Cell-attached and patch-clamp recordings were recorded in the same session. Electrodes with resistances of 4 to 8 MΩ were pulled from thick-walled borosilicate glass capillaries and filled with (in mM): 9 KCl, 10 KOH, 3.48 MGCl 2 , 4 NaCl, 120 K-Gluconate, 10 Hepes, 28.5 Sucrose, 4 Na 2 ATP and 0.4 Na 3 GTP. An Axopatch 700B amplifier (Molecular Devices) was used for recordings. Data were filtered at 1 kHz using pClamp software and with offline analysis performed with Clampfit 10.5 (Molecular Devices). All recordings were performed between 13 and 18 DIV.
Neurons were recorded in whole-cell configuration with the same solutions, pipettes, equipment and analysis software as described above. Membrane test settings were recorded in voltage clamp with the holding potential at −65 mV. We recorded the resting membrane potential in current clamp without a bias current directly after breaking the giga-seal. This potential was corrected for the liquid junction potential (−10.5 mV).

NPC injections in early postnatal and adult cerebella. Cell injections were performed on eight P1
mice. P1 mice were cryo-anesthetized, and the cerebellum was localized by trans-illumination of the head. A single cell suspension of NPCs, at a concentration of 5 × 10 4 cells/µl in NS21 medium and labeled with Fast Green, was injected by means of a glass micropipette with a tip diameter of around 60-80 µm and a pneumatic pressure injection apparatus (Picospritzer II, General Valve, Fairfield). About 1 µl of cell suspension was administered at each injection site, which included the vermis and each hemisphere of the cerebellum.
Eight 4-month-old and eight 10-month-old wild type animals from different litters, three 5-month-old cyclin-dependent-kinase-9 (CDK9)-mCherry knockin 50  L7cre-ERCC1 27 homozygous mice were anesthetized through inhalation of 5% isoflurane in O 2 for 2 hrs. A single cell suspension at a concentration of 5 × 10 4 cells/µl in NS21 medium was obtained from cultured NPCs. Wild type and homozygous L7Cre-ERCC1 KO mice received three injections (1.5 µl total volume at 0.5 µl/min) of the final cell suspension using a stereotaxic apparatus with a Hamilton syringe (Hamilton, Germany) according to the following coordinates from Lambda targeting the deep cerebellar nuclei (DCN), lobule 4/5, simple lobule and crus1 of the cerebellum, respectively: Immunohistochemistry. EBs were collected from 11 to 15 DIV in 96-well plates and concentrated in 50 ml tubes. Subsequently these were centrifuged at 300 g for 5 min and the supernatant was removed. EBs were then re-suspended in paraformaldehyde (PFA, 4%), transferred to a 1.5 ml tube and stored overnight at 4 °C. EBs in 1.5 ml tubes were washed twice with phosphate buffer after centrifugation, followed by careful removal of the supernatant. Immediately after EBs were embedded in Tissue-Tek (Sakura Finetek Europe, The Netherlands) and stored at −80 °C. Sections (10 µm) of EBs were obtained using a Cryotome (Microm HM 560, Microm Gmbh, Germany) stained on glass overnight at 4 °C with primary antibodies and for 2 h with fluorescent secondary antibodies (see below for antibodies and dilutions).
Neurons on coverslips were fixed (4% PFA) for 15 min at 4 °C and treated with primary and secondary antibodies as described above. Neurons/Matrigel in µ-wells were incubated for 2 nights at 4 °C with primary antibodies and overnight at RT with fluorescent secondary antibodies. To prevent the Matrigel from disintegrating µ-well antibody stainings were performed in a solution containing: 0.05 M Tris buffer, 0.9% NaCl, 0.25% gelatine, and 0.5% Triton-X-100 at pH 4. For visualization of recorded neurons, biocytin was added to the intracellular solution, incubated in 4% PFA for two nights and treated with streptavidin-Cy3.
Free floating sections were processed with antibodies as described above for EBs sections. For quantitative analyses, at least 3 images from over 3 experiments were evaluated. The following antibodies were used at the Statistical Analysis. The mean and deviation of measurements were evaluated in Matlab (Mathworks) using one-way ANOVAs followed by Tukey's post-hoc testing with statistical significance defined at p < 0.01, unless stated otherwise. All data are reported as mean ± standard error of measurements (SEM).