Sonic Hedgehog Agonist Protects Against Complex Neonatal Cerebellar Injury

All animal procedures were reviewed and approved by the Institutional Animal Care and Use Committee and Laboratory Animal Resource Center at University of California, San Francisco (UCSF). Mouse colonies were maintained in accordance with NIH and UCSF guidelines. C57Bl6/J mouse lines were obtained from the Jackson Laboratories. Gli-luciferase [49], VHL floxed [50], Atoh1-cre (Math1-cre) [51], and L7-cre [19] mice have been previously described.

Synthesis of SAG has been previously described [52]. SAG was dissolved in dimethyl sulfoxide (DMSO) to 5 mM and further diluted with normal saline or culture medium. These experiments used SAG as a free-base form. Vehicle controls comprised saline containing an equivalent concentration of SAG.

Chronic hypoxic rearing was performed as previously described [35, 48, 53, 54]. Briefly, litters of the C57Bl6/J strain were culled to a size of, at most, ten pups and co-fostered with CD1 or Swiss Webster strain dams then reared at 10% O₂ in a hypoxic chamber (Biospherix, Inc., Laconia, NY) from postnatal day 3 (P3) to P11. Tissue from P4, P7, P11, P22, or P40 was then harvested acutely for analysis. On P3, pups received daily intraperitoneal injections of prednisolone (0.67 g/kg body weight, Sigma-Aldrich), SAG (20 g/kg body weight), prednisolone in combination with SAG, or vehicle (DMSO), for 8 days, or the duration of the hypoxic experiment. For acute treatment, a one-time dose of SAG was given at P11.

Following intracardial perfusion of 4% paraformaldehyde, tissue was post-fixed for 1–2 h at 4 °C, cryoprotected in 30% sucrose, and embedded in OCT. Frozen sections were cut on a cryostat (20 μm) and stored at −80 °C. For staining, sections were thawed and then rehydrated in PBS. If necessary, tissues underwent antigen retrieval with citrate buffer (pH 6.0) at 95 °C for 10 min. Sections were blocked in 10% goat or donkey serum in 0.1% TritonX-100-containing PBS (blocking solution). Primary antibodies were diluted in blocking solution, and tissues incubated at 4 °C overnight or 2-h room temperature. For primary antibodies, we used PH3 (mouse monoclonal, Cell Signaling), Calbindin (mouse monoclonal or rabbit polyclonal, Swant), Cleaved Caspase 3 (rabbit polyclonal, Cell Signaling), NeuN (mouse monoclonal, Millipore), Iba1 (rabbit polyclonal, Wako), HIF1a (rabbit polyclonal, Cayman Chemicals), BNIP3 (rabbit polyclonal, Cell Signaling), and Cre (rabbit polyclonal, Millipore). Following primary incubation, tissues were washed with 0.1% Tween20-containing PBS, then incubated with proper Alexa Fluor secondary antibodies (Invitrogen) in blocking solution for 1 h at room temperature. Sections were mounted with fluoromount containing DAPI (Southern Biotech).

Primary cultures were performed as previously described [30]. Briefly, either C57Bl6/J or Gli-luciferase pups were euthanized at P4 or P5, and their cerebella were dissected and dissociated. Cultures were maintained in serum-free medium containing only vehicle or ShhN for 24 h prior to treatments. Cultures were then transfected with a HIF1α overexpressing plasmid or an empty vector for 12 h using the Piggyback transposon system to allow high efficiency expression as previously described [55]. Cells were collected 12 h afterward for protein extract or luciferase activity using the Dual-Luciferase Reporter Assay System (Promega). For analysis of Shh targets under hypoxic conditions, cultures were incubated in a 1% O₂ incubator for 24 h, or treated with dimethyloxalylglycine (DMOG) (Sigma), a PHD4 inhibitor that upregulates HIF.

Preparation of protein extracts, immunoblots, and fluorescent detection was done as previously described using the Li-Cor Odyssey system (Li-Cor, Lincoln, NE) [34]. Antibodies used include HIF1α (rabbit polyclonal, Cayman Chemicals), Cyclin D1 (rabbit polyclonal, ThermoScientific), Patched 1 (goat polyclonal, Santa Cruz), Gli1 (rabbit polyclonal, Santa Cruz), Gli3 (goat polyclonal, R and D Systems), N Myc (mouse monoclonal, Millipore), and β-Actin (mouse ascites, Sigma).

To assess Gli1 mRNA expression in the cerebellum, we used the RNAscope LS Multiplex Assay (Advanced Cell Diagnostics). P11 mouse brain sections were baked at 65 °C for 45 min, post-fixed in 4% paraformaldehyde for 15 min, and dehydrated prior to the assay. Multiplexed single-molecule fluorescent in situ hybridization (smFISH) against Gli1 and Atoh1 mRNA was performed on the BOND RX automated stainer (Leica) using the RNAscope reagents. The Gli1 probe consists of 20 double-Z probes targeting nucleotides 25–1025 on the mouse transcript NM_010296.2 and was developed with the fluorophore Opal 520 (Perkin Elmer). The Atoh1 probe consists of 20 double-Z probes targeting nucleotides 847–2088 on the mouse transcript NM_007500.4 and was developed in Opal 570.

Surface area of the entire cerebellum was calculated as previously described [30]. Briefly, the cerebellum from immunostained sections with DAPI was outlined, and surface area was measured in Adobe Photoshop CS6 using the Record Measurements function, with measurement scale set to match the objective lens from the microscope (e.g., for the Zeiss × 10 objective, 1 pixel = 0.65 μm). For area measurement and cell counting of immunohistochemistry (IHC) from sagittal brain sections of P4, P7, P11, and P22 wild-type and transgenic animals, tiled images comprising the entire cerebellar vermis were taken with a Zeiss AxioImager.Z2 microscope equipped with a motorized stage using either the × 10 or × 20 objective. The mean area per section for each animal was determined in mm² from measurements of five parasagittal sections from the midline region of the cerebellum, and is used as a proxy for volume. Quantification of cerebellar cross-sectional area resulted from an average of n = 3 animals in normoxic and hypoxic conditions for P4 and P22; n = 3 in normoxic, and n = 4 in hypoxic conditions at P40. For the hypoxia + prednisolone experiments, quantification of the cerebellar vermis is resulted from n = 3 animals for the Nx and Hx conditions, and n = 5 animals for the Hx + prednisolone and Hx + prednisolone + SAG conditions. For experiments using transgenic animals, quantification of the cerebellar IGL resulted from n = 3 animals for the Math1Cre;Vhl(fl/+) and Math1Cre;Vhl(fl/fl) conditions, n = 7 for Math1Cre;Vhl(fl/+) + prednisolone condition, and n = 6 for the Math1Cre;Vhl(fl/fl) + prednisolone condition; n = 3 for L7Cre;Vhl(fl/+), L7Cre;Vhl(fl/fl), and L7Cre;Vhl(fl/+) + prednisolone conditions, and n = 4 for L7Cre;Vhl(fl/fl) + prednisolone condition. The area of the cerebellum was measured using either Adobe Photoshop CS6 or Stereo Investigator (MBF Biosciences).

The numbers of immunopositive or double-immunopositive cells were quantified by a blinded investigator using the Count function on Adobe Photoshop CS6, or the Spot Detection function on Imaris (Bitplane). The average cell count for each animal was determined from measurements of five to seven whole parasagittal sections from the midline region of the cerebellum. Quantification of Calbindin-positive cells resulted from n = 3 animals in normoxic and hypoxic conditions, n = 6 in the hypoxia + prednisolone condition, and n = 5 in the hypoxia + prednisolone + SAG condition.

For all quantified data, mean + SEM values are presented. Statistical analysis was performed using unpaired, two-tailed Student’s t tests, and with an ANOVA (single factor). For a significant difference (p < 0.05), a Tukey’s post-hoc test was performed (GraphPad Prism).

To study the effect of low oxygen tension on development of the brain, wild-type C57B6/J females with new litters were placed in a hypoxic (10% FiO₂) cage for 7 days from postnatal days (P) 3–11 (Fig. 1a), a model that has previously been used to study reproducible phenotypes, including delayed myelination [34, 47]. After 24 h of hypoxia, where pups were reared in the chamber from P3 to P4, we observed reduced cerebellar size that was significant (Fig. 1b, c; P4 Nx vs. Hx, p = 0.0432; EGL outlined for clarity) despite lower body weight of hypoxic pups (Table 1). It has previously been shown that the cerebellum is disproportionately affected by glucocorticoid treatment compared to brain and body size [30]. Similarly, the cerebellum is disproportionately affected under present conditions (Table 1). However, at P21 and P40, hypoxia-exposed mice showed ongoing cerebellar hypoplasia, despite catch-up in body weight and cerebral size (Table 1), with non-hypoxia-exposed littermates (Fig. 1b, c; P40 Nx vs. Hx, p = 0.018; IGL outlined for clarity).

Cerebellar hypoplasia could possibly be due to cell death or decreased proliferation in the external granule layer (EGL). We first analyzed cleaved caspase 3 (Casp3) as a marker of apoptosis and phospho-histone H3 (pH 3) as a marker of mitosis. While there was no significant difference for Casp3 counts between hypoxic and normoxic mice (Fig. 2c), pH 3 counts were significantly decreased in CGNPs of at P11 (Fig. 1d, e; Nx vs. Hx, p = 0.002). As CGNP proliferation is driven by Shh signaling, we investigated whether decreased proliferation was a result of Shh downregulation. As shown (Fig. 1f, g), we found that levels of the Shh targets Patched 1 (Ptch1), N-myc, and cyclin D1 (CCD1) were downregulated in whole cerebella lysates from P11 hypoxic pups. To confirm downregulation of Shh signaling specifically in CGNPs, we collected protein lysates from primary cultures incubated in a 1% O₂ hypoxic incubator for 24 h, or cultures treated with DMOG to upregulate HIF signaling (Supplementary Fig. 1). We found decreased levels of Ptch1 and increased levels of Gli3, which act as a repressor of Shh signaling, under DMOG or hypoxic conditions compared to control conditions. Treating our cultures with ShhN, which mimic the proliferating EGL in vivo, increased target protein levels of Gli1, N-myc, and Ptch1 as expected.

We have previously shown that Sonic hedgehog-Smoothened agonist (SAG) administration in postnatal pups crosses the blood–brain barrier and upregulates a Shh signaling (Gli-luciferase) reporter in the cerebellum [29]. As shown (Fig. 1e–g), SAG given daily from P3–11 resulted in a significant increase in cells in mitosis in the EGL only under hypoxic conditions (Fig. 1e; Hx vs. Hx + SAG, p = 0.002; Nx vs. Nx + SAG, n.s.) and Shh target protein levels from whole cerebellar lysates (Fig. 1f, g). Although chronic hypoxia may affect many cerebellar populations to contribute to hypoplasia, these findings suggest that hypoxia causes cerebellar hypoplasia in part through downregulated Shh pathway activity in CGNPs that can lead to decreased proliferation.

Human preterm infants with chronic lung disease may become exposed to intermittent hypoxemia and postnatal glucocorticoids [7, 29]; thus, we investigated the combined impact of hypoxic rearing and GC administration on cerebellar development (Fig. 1a). We used prednisolone based on our previous finding that it causes cerebellar injury and can be rescued by SAG administration, which promotes upregulation of the protective enzyme 11βHSD2 in CGNP [29]. We first investigated Shh signaling under injury conditions specifically in the EGL and CGNPs. Fluorescent in situ hybridization shows that the Shh downstream target Gli1 is reduced in P11 cerebella in CGNPs, which express Atoh1 (Fig. 2a, left and center images) in both chronic hypoxia (Hx) and chronic hypoxia combined with daily prednisolone (Hx + Pred) administration from P3 to 11. This could be due to gene downregulation or reduced population of CGNPs. Unexpectedly, we observed ectopic expression of Gli1 level in the Purkinje cell layer (PL, Fig. 2a, right; Supplementary Fig. 4) only in the condition of dual injury (Hx + Pred), which might be attributable to Gli1 expression in either Bergmann glia or invading microglia (described below and in discussion).

As shown (Fig. 2b, d), we found increased apoptosis via Cleaved Caspase 3 (Casp3) staining in the P11 cerebellum of animals subjected to hypoxia plus prednisolone (Nx vs. Hx, n.s.; Nx vs. Hx + Pred, p < 0.001). Apoptotic cells included Calbindin + Purkinje cells and NeuN + cerebellar granule neurons (CGNs) of the internal granule layer that together comprised over half of the total Casp3+ cells (Fig. 2b–e). Additionally, we observed abnormalities in the Purkinje cell layer, including arborization defects and cell soma enlargement (Fig. 2b, f, Supplementary Figs. 2 and 3). In posterior lobules 7–9, we observed areas of markedly depleted Calbindin signal indicative of massive Purkinje cell loss (Supplementary Fig. 2). Inflammation was greater in the dual injury model, which showed an increase in activated Iba1+ microglia that invaded into the Purkinje layer, with colocalization of Iba1 and Calbindin noted in a few Purkinje cells (Fig. 2f and insert, g; Hx vs. Hx + Pred, p < 0.05). It is not clear whether increased Iba1+ microglia result from the dual insults themselves or increased cell death and activation to clear debris. In contrast, animals exposed to hypoxia alone showed fewer than 5% of neuronal cells that were Casp3+ (Fig. 2e). Together, these data indicate that glucocorticoid administration plus hypoxia exacerbates cerebellar injury targeting both CGNs and Purkinje cells.

SAG is neuroprotective for the neonatal cerebellum in mouse models of glucocorticoid injury [29] and Down syndrome [56]. We next investigated whether SAG was beneficial in the setting of complex injury using two administration paradigms (Fig. 1a). In the first procedure, SAG (20 mg/kg body weight) was administered daily with prednisolone from P3 to P11. In the second case, we gave only a one-time dose of SAG (20 mg/kg body weight) after injury at P11. In both cases, we analyzed the cerebellum at P22. As shown (Fig. 3a, b), pups reared under hypoxia + prednisolone conditions showed greater vermian cerebellar hypoplasia and tissue damage, notably a reduction in the number of lobules, than littermates subjected to hypoxia only (p < 0.001). However, hypoplasia was significantly improved with SAG treatment (p < 0.01). As shown (Fig. 3a, b), SAG given as either in a chronic regimen or an acute one-time administration was effective at promoting partial rescue of cerebellar volume.

In addition, we found that SAG protected against Purkinje cell death. Purkinje cell survival after hypoxia + prednisolone was significantly (p < 0.001) compromised at P22 compared to animals that received hypoxia alone (Fig. 3c, d), which was especially prominent in posterior cerebellar lobules 6–10 (Fig. 3a, c). In addition to a decrease in Purkinje cell number, hypoxia + prednisolone caused a disruption in the Purkinje layer and dendritic arborization that is visible at P11 (Figs. 2f and 3e, middle panel insert) and began between P7 and P9 (Supplementary Fig. 3). By P22, cleaved caspase 3-positive cells were no longer detectable in the Purkinje layer. As shown (Fig. 3c, d), one-time SAG administration was able to partially rescue Purkinje cell loss (p < 0.01) and many of the surviving Purkinje cells showed a more normalized and less hypertrophic morphology (Fig. 3e, f). In contrast, SAG administration did not appear to rescue dendrite morphology (Fig. 3e).

These results demonstrate that hypoxia + prednisolone administration results in complex cerebellar injury, and that even a one-time dose of SAG at P11 post-injury partially protects the cerebellum from adverse impact. It seems likely normalization of CGNP proliferation with SAG is via the canonical Shh pathway [29, 30] (Fig. 1f, g), whereas the mechanisms resulting in Purkinje cell survival are less clear and require further study. It is possible that abnormalities are in part secondary to the decrease in CGNP cells as previously reported [19], while the loss of Shh activity resulting from hypoxia and prednisolone may influence late Purkinje cell maturation.

We next investigated cell type-specific effects of HIF function plus Pred in conditional mutant transgenic mice. HIF1α is targeted for proteosomal degradation by the von Hippel Lindau (VHL) factor under normoxic conditions (Fig. 4a) [42, 43]. We first targeted CGNPs for HIF activation using Math1Cre;Vhl(f/f) transgenic mice (Fig. 4b; EGL layer in red). As expected, in normoxic mice, CGNPs showed strong HIF1α expression at P2. Furthermore, BNIP3, a target of HIF activation, was upregulated specifically in the IGL at P2 and P11 (Fig. 4c, data not shown), demonstrating HIF transcriptional activity. However, because the phenotype does not capture the full impact of hypoxia in CGNP, we conclude that other hypoxia-induced factors besides HIF or HIF signaling in other cells accounts for the full impact of hypoxia. HIF overactivation did not cause cerebellar hypoplasia, although there was a trend toward decreased cerebellar size at P2 that did not reach significance (Fig. 4d). HIF activation resulted in decreased CGNP proliferation at P2 (Fig. 4b insert, e; p < 0.05), which eventually normalized at later ages (data not shown). Basal BNIP3 expression in the Purkinje layer was also observed in both wild-type and transgenic conditions.

Although high HIF expression alone did not cause cerebellar hypoplasia, we hypothesized that HIF primed CGNPs for glucocorticoid-mediated injury, as previous reports link HIF activation with upregulation of the glucocorticoid receptor (GR) [57], and there can be crosstalk between hypoxia-dependent signals and glucocorticoid-mediated gene regulation [58]. Indeed, administration of prednisolone from P3 to 11 in Math1Cre;Vhl(fl/fl) animals caused significant (p < 0.01) decrease in the cross-sectional area in the vermis (Fig. 4f, g). These findings indicate that prednisolone administration combined with CGNP-specific HIF activation results in cerebellar hypoplasia.

VHL has other functions besides HIF1α degradation [59]. To further investigate HIF1α activity per se, we cultured CGNPs and transfected them with an overexpressing construct. Protein analysis revealed that HIF1α expression was greater in transfected conditions, and that the G1 cell cycle protein Cyclin D1 was downregulated (Fig. 4h). By using Gli-luciferase reporter mice [49], we found that transfected cultures showed lesser Shh activity in CGNPs (Fig. 4i). We conclude that HIF1α activity inhibits CGNP Shh activity and proliferation, and primes CGNPs for enhanced toxicity when exposed to glucocorticoids.

The Purkinje cell is generally thought to be highly vulnerable to hypoxic injury in humans [3] and rodents [60]. Thus, we predicted that HIF1α stabilization in Purkinje cell-specific Protein 2 (Pcp2/L7)-cre;Vhl(fl/fl) animals would have devastating effects (Fig. 5a). In these animals, L7 cre activity commences before birth and gradually becomes expressed in all Purkinje cells. As shown (Fig. 5b), cre expression is first detected in caudal cerebellum and becomes generally expressed in Purkinje cells by P8. As expected, HIF1α expression was specific to Purkinje cells in L7cre;Vhl(fl/fl) animals at P22 (Fig. 5c).

Surprisingly, we saw no evidence of Purkinje cell loss in L7cre;Vhl(fl/fl) animals despite strong upregulation of HIF1α. However, when compounded with prednisolone administration from P3 to 11, analysis at P22 revealed significant (p = 0.0079) Purkinje cell losses (Fig. 5d, e). We note that the caudal cerebellar lobules were the most affected, consistent with earlier cre driver activity in this region (Fig. 5b). Despite this reduction in Purkinje cell number, cerebellar size was unaffected (L7Cre;Vhl(fl/+) = 5.25 ± 0.037 mm², n = 6; L7Cre;Vhl(fl/fl) = 5.20 ± 0.146 mm², n = 4, n.s.; data not shown); further studies are needed to determine timing of death. Thus, HIF activation primes serious Purkinje cell injury when they are exposed to glucocorticoids. Together, these transgenic animal data indicate a role for HIF signaling to potentiate glucocorticoid-mediated injury in two neuronal populations of neonatal mouse cerebellum (Fig. 6).

Chronic lung disease, which can cause intermittent hypoxemia, is associated with cerebellar hypoplasia in human infants. We observed that chronic hypoxia caused cerebellar hypoplasia, which likely reflects inhibition of Shh pathway and decreased proliferation in CGNPs. Cerebellar size differences were evident as early as 24-h post-exposure to hypoxia in mice, with the difference becoming greater after 7 days. This cerebellar hypoplasia was evident at P40 demonstrating permanent damage. In contrast to CGNP, Purkinje cell numbers were unaffected by hypoxia alone.

The Shh pathway drives CGNP proliferation [12, 15‐17, 19], and the Shh target genes N-myc, Ptch1, and CCD1 were downregulated when postnatal pups were subjected to hypoxia. This may be due to the HIF-pathway, since primary CGNP cultures overexpressing HIF resulted in decreased CCD1 and Gli1 transcription and decreased proliferation. In vivo, we observed a correlate at P2 with decreased CGNP proliferation in Math1Cre;Vhl(fl/fl) pups; however, this normalized at later stages and these animals did not have permanent cerebellar hypoplasia. These findings suggest that Shh is likely the target of hypoxia-induced cerebellar hypoplasia, and that a compensatory mechanism might operate in vivo to counteract pure HIF1α stabilization in normoxic pups after P2.

Postnatal glucocorticoids such as dexamethasone, prednisolone, and corticosterone are used to treat hypotension and chronic lung disease in the NICU [7, 61, 62]. When we combined hypoxic rearing with prednisolone administration, we observed exaggerated cerebellar hypoplasia at P22 and significant Purkinje cell death compared to either condition alone. We used Math1Cre;Vhl(fl/fl) or L7Cre;Vhl(fl/fl) transgenic mice to specifically target VHL loss-of-function in CGNP or Purkinje cells, respectively, to better understand how HIF signaling and glucocorticoids might act in cell type-specific ways. While Math1Cre;Vhl(fl/fl) did not show long-term cerebellar abnormalities, the addition of prednisolone caused cerebellar hypoplasia with long-term consequences to the IGL. Prednisolone in Math1Cre;Vhl(fl/+) or wild-type animals did not result in this reduction; this is in contrast to previous work [30], possibly due to delayed administration of prednisolone, starting at P3 instead of P0. Thus the impact of increased HIF signaling in CGNPs may be priming the cerebellum for a second insult, in our case, prednisolone. We found expression of the HIF target BNIP3 in Math1Cre;Vhl(fl/fl) animals at P11 in the EGL but not IGL (Fig. 4c). Thus, proliferation-related effects of HIF on CGNPs most likely occur at earlier timepoints. This suggests that there are two different threshold points for prednisolone-induced injury in our transgenic models. We observe early HIF activation in the EGL before P11, and a continuous HIF signaling in the Purkinje cells that occurs past P11, as seen by BNIP3 in the Purkinje cells. In L7Cre;Vhl(fl/fl) animals, prednisolone administration caused cell death of the Purkinje cell population by P22. How the HIF-pathway contributes to glucocorticoid-mediated exaggerated injury remains unclear but may be related to activation of glucocorticoid receptor signaling by HIF [57, 64‐66]. Our experiments showed a strong additive effect of hypoxia and prednisolone on degree of injury, dependent on the HIF pathway (Fig. 6).

Based on previous studies showing that Smoothened agonist (SAG) can sustain cerebellar growth in the face of exogenous insults and in a Down syndrome mouse model [29, 56], we tested SAG in the setting of hypoxia-prednisolone combinatorial injury. SAG is effective in preventing glucocorticoid-induced cerebellar neurotoxicity in neonatal mice in an 11βHSD2-dependent manner [29, 30]. Prednisolone is an 11βHSD2 sensitive glucocorticoid and is the reason we chose to use this compound versus dexamethasone, which has been used in other studies of glucocorticoid-mediated injury [67‐71]. Because SAG is a potent activator of the Shh pathway [52], there is theoretical risk of tumor formation (e.g., medulloblastoma) [72, 73]. However, previous studies are reassuring and show that SAG is non-tumorigenic [29]; animals studied here also showed no ill effects from SAG with normal gains in body weight.

When we administered SAG either chronically from P3 to 11 or as a one-time dose at P11, we saw increased CGNP proliferation and normalized cerebellar volume despite prednisolone + hypoxia exposure. The beneficial effects of SAG in CGNPs may be due to inactivation of prednisolone itself and potentially higher levels of Shh pathway activation, which would act to support CGNP proliferation. In a Down syndrome mouse model (Ts65Dn), one-time SAG administration rescued both cerebellar morphology and phenotypes associated with hippocampal deficits [56].

Surprisingly, SAG was also able to rescue Purkinje cell loss, although the protective mechanism is unclear. Purkinje cells appeared hypertrophic at P22 following hypoxia and GC administration. With SAG administration, Purkinje cell numbers were normalized and they did not show hypertrophy. We found arborization defects from Purkinje cells starting between P7 and P9, and reduced Purkinje cell number by P22. Zonouzi et al. [47] previously reported that chronic hypoxia resulted in reduced arborization by P7 without reducing Purkinje cell number. It is possible that the 2-day delay in aberrant arborization in our experiments is due to genetic strain differences or different markers: in the case of Zonouzi et al., they used transgenic NG2DsRed or GAD65-GFP mice of mixed genetic background. We observed ectopic Gli1 upregulation in the Purkinje layer that was specific to the case of hypoxia plus prednisolone. Although further studies are needed to confirm which cell population is expressing Gli1, many Gli transcripts are observed outside the borders of Calbindin + Purkinje cells, suggesting the possibility that ectopic Gli1 expression occurs in Bergmann glia or invading microglia under these conditions. In any case, SAG might rescue Purkinje cell death via trophic effects on dendrites. Alternatively, SAG might protect the cerebellum via anti-inflammatory roles and potentially promoting blood-brain barrier (BBB) integrity [74‐76]. It is also possible that Bergmann glia are benefitted by SAG or that it might protect Purkinje dendrites in the molecular layer. Indeed, a recent study shows that Shh signaling is necessary to maintain normal Bergman glial phenotype [77].