Engrailed2 modulates cerebellar granule neuron precursor proliferation, differentiation and insulin-like growth factor 1 signaling during postnatal development

Time-mated Sprague-Dawley rats were obtained from Hilltop Lab Animals, Inc. (Scottdale, PA) and maintained on a 12:12 light:dark cycle with ad libitum Purina rat chow and water. Rats were killed by CO₂ gas asphyxiation as approved by RWJMS IACUC. Male and female breeding pairs of C57BL6 J/129S2SV PAS mice heterozygous (HT) for a functional deletion of En2[13] were obtained from The Jackson Laboratory (no. 002657; Bar Harbor, ME) and maintained. An En2 mutant colony was propagated by HT × HT intercrosses, and genotyping of the progeny was performed as described (jaxmice.jax.org). The mice were initially maintained as heterozygous matings, but, to generate adequate numbers of pups of known genotype for GNP cultures, WT × WT and KO × KO mating pairs were established.

All animal procedures were assessed and approved by the Robert Wood Johnson Medical School Institutional Animal Care and Use Committee. Animals were managed by Robert Wood Johnson Animal Facility, and maintenance, husbandry, transportation, housing and use were in compliance with the Laboratory Animal Welfare Act (PL 89–544; PL-91-579) and National Institutes of Health guidelines (Manual Chapter 4206).

A full-length En2 cDNA was cloned from C57BL6 J/129S2SV PAS adult mouse cerebellum as described previously [7]. To identify transfected cells, the En2 coding sequence was moved to the pCMS-EGFP expression vector (Clontech, Mountain View, CA) by EcoRI digestion and ligation. Orientation of positive clones was determined by Kpn1 digestion.

Postnatal day (P)7 rat and WT or KO mice were decapitated and cerebellar GNPs isolated as described previously [37, 38]. Briefly, 3–6 cleaned cerebella were incubated 3 min in trypsin-DNase solution (1% trypsin, 0.1% DNase, Worthington, Lakewood, NJ) and dissociated in DNase solution (0.05% in DMEM) by trituration. After pelleting, cells were filtered (30-u m nylon mesh; Tekton, Tarrytown, NY), resuspended and centrifuged at 3,200 rpm on a Percoll (Sigma, St. Louis, MO) 35:60% step gradient [39, 40]. Cells at the 35:60% interface were collected and washed in phosphate buffer, then plated onto a poly-D-lysine (0.1 mg/ml)-coated 60-mm culture dish in defined medium (DM) composed of a 1:1 mixture of F12 and DMEM, 10 ng/ml insulin, 100 u g/ml transferrin, 10 mg/ml bovine serum albumin, 100 u M putrescine, 20 nM progesterone, 30 nM selenium, 6 mg/ml glucose, 50 U/ml penicillin and 50 u g/ml streptomycin. Unless stated otherwise, components were obtained from Sigma. After 1 h of preplating to remove adherent flat cells (<2% of cells), small round (granule) cells were dislodged by gentle pipetting and plated at ~10⁴-5 × 10⁴/cm² on poly-D-lysine-coated 35-mm Nunc (Thermo Fischer Scientific, Rochester, NY) culture dishes or 24 multiwells. Cultures were maintained in a humidified 5% CO₂/air incubator at 37°C.

For thymidine incorporation studies, cells were cultured in DM alone or with additional factors: high-dose insulin (10 u g/ml), 1–100 ng/ml IGF1 (Cell Sciences, Canton, MA), 0.1-3 u g/ml Shh-N (N-terminal fragment, R&D Systems, Minneapolis, MN), 10^-12-10^-7 M PACAP1-38 in 0.01 N acetic acid vehicle (American Peptide, Sunnyvale, CA), 10 ng/ml FGF2 (gift of Scios, Mountain View, CA), 10–30 ng/ml BDNF (PeproTech, Rocky Hill, NJ), 100 ng/ml EGF, 100–300 ng/ml VEGF or 10-30% conditioned media (CM) harvested from control- or Wnt3a-transfected fibroblasts.

For differentiation, cells were assessed live 24 h post plating by phase microscopy or were fixed in 4% cold paraformaldehyde and processed for immunocytochemistry, then assessed by phase or fluorescence microscopy. To assess neurites, cells bearing processes greater than two-cell soma diameters were assessed at 24 h in the live state, as reported previously [41], or following fixation and immunostaining against beta-III-tubulin as described below. Cells in two or three 1-cm rows (1.0-1.5% of the culture dish area) bearing neurites >2 cell soma were counted blind in two to four dishes per group, and experiments were performed three or more times.

Incorporation of ³H-deoxythymidine (³H-dT, 1u Ci/ml) was used to assess DNA synthesis, as described previously [38, 42, 43]. Cells were incubated with ³H-dT during the final 4 h of culture. DNA that had incorporated ³H-dT was collected onto glass fiber filters by a semiautomatic cell harvester (Skatron), and the radioligand was assessed by scintillation spectroscopy. To visualize cells synthesizing DNA, GNPs were exposed to the S-phase marker bromodeoxyuridine (BrdU) (10 u M, Sigma) during the final 4 h of incubation. After fixation, cells were either exposed to 2 N HCl (30 min) or pretreated with PBS/0.3% Triton X-100 and then DNaseI (500 U/ml, Worthington) [44] and subsequently processed for BrdU immunocytochemistry using monoclonal anti-BrdU (1:200; Dako, Carpinteria, CA) and visualized using a Vectastain avidin-biotin complex kit and Vector SG peroxidase substrate with a 3,3′-diaminobenzidine (DAB) chromogen (Vector Laboratories, Burlingame, CA) as previously described [45]. The labeling index, defined as the proportion of total cells incorporating BrdU into the nucleus, was determined by scoring the cells in five randomly selected, non-overlapping fields in each of the two to four dishes per group per experiment.

Rat and mouse GNPs were plated 2-4 × 10⁵cells/cm² onto 12- or 25-mm glass coverslips (VWR International, West Chester, PA) pretreated with 2 N HCl for 30 min, rinsed in dH₂O for 30 min, serially washed in 90% and 100% ethanol and fire-polished, then coated with poly-D-lysine (0.1 mg/ml) and fibronectin (1 u g/cm², Sigma). Transfection media consisted of Neurobasal supplemented with 2% B27 (Invitrogen) containing glutamine (Gln, 2 mM), penicillin (50 U/ml), streptomycin (50 mg/ml), BSA (1 mg/ml), and either FGF2 (10 ng/ml) for rat GNPs or Shh (3 u g/ml) and BDNF (30 ng/ml) for mouse GNPs. For both species, 1 h after plating, cells were gently washed with Neurobasal supplemented with 10 u g/ml Gln, then incubated with transfection reaction media containing Neurobasal supplemented with 2% B27 and 10 u g/ml Gln, and either Lipofectamine Plus or Lipofectamine LTX Plus (Invitrogen) and 1 u g/ml pCMS-EGFP (GFP) or 1.2 u g/ml pCMS-EGFP-En2 (En2). After 5 h, transfection reaction media were replaced with species-specific transfection media (as above) and cells were incubated an additional 24 h, then fixed in cold 4% paraformaldehyde. Successful transfection was determined by GFP autofluorescence, assessed by fluorescence microscopy prior to fixation or other procedures.

Following fixation, cells were incubated 1 h at room temperature (RT) in 33% goat, rabbit or horse serum, then 1 h at RT or overnight at 4°C in one of the following primary antibodies diluted in PBS/0.3% Triton X-100, 2% serum and 0.05% NaN₃: TuJ1, monoclonal mouse anti-beta-III-tubulin (1:9,000, Covance Research, Inc., Berkeley, CA), monoclonal mouse anti-PCNA (F2) (1:1,000, Santa Cruz, Inc., Santa Cruz, CA) or polyclonal chicken anti-GFP (1:5,000, Chemicon). Cells were then incubated 1 h with biotinylated horse anti-mouse secondary antibody and visualized with DAB chromogen or, alternatively, Alexafluor 488 (green) or 594 (red) fluorescent goat anti-mouse, goat anti-rabbit, rabbit anti-chicken, rabbit anti-mouse or rabbit anti-goat secondary antibody (1:1,000, Chemicon), followed by 5 min incubation with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI, 1 u g/ml, Sigma). Glass coverslips were inverted and mounted onto glass slides with Fluoromount-G (EMS, Hatfield, PA) and assessed by fluorescence microscopy. Cells were counted blind from two randomly selected fields within each of ten predetermined regions covering greater than 80% of the coverslip, from 2–4 coverslips per group per experiment. Images were acquired at 400× with and without background subtraction via an Apotome filter using Axiovision 4.5 software (Carl Zeiss Microimaging, Inc., Thornwood, NY).

Isolated P7 KO and WT GNPs (2.5 × 10⁶ cells/ml) were incubated in 4 ml DM on poly-D-lysine-coated (0.01 mg/ml) 60-mm dishes. After 4 h incubation, 10 ng/ml IGF1 was administered, followed by 15 min to 8 h incubation, then cells were washed with PBS, lysed by 70 u l lysis buffer containing 20 mM HEPES-KOH, pH 7.5, 5 mM KCl, 0.5 mM MgCl₂ with 0.5 mM dithiothretiol (DTT), 2 mM phenylmethylsulfonylfluoride (PMSF), 1 mM leupeptin and 3.5 mg/ml aprotinin, and collected by scraping with rubber policeman. Samples were stored at -80°C. For blotting, defrosted samples were lysed by sonication and centrifuged, and the supernatant was normalized to 0.1 M NaCl. Protein concentrations were determined by Bio-Rad protein assay (Bio Rad Laboratories, Inc., Hercules, CA), and bovine serum albumin (BSA) dilutions served as standard curve. Protein extracts (50 u g per lane) were analyzed by 12% SDS-PAGE and electrically transferred to a polyvinylidenediflouride (PVDF) membrane. The membrane was blocked with 5% milk or 5% BSA and incubated with primary antibody against phospho- or total Akt, phospho- or total ERK1/2, phospho- or total GSK3beta, phospho- or total S6 kinase (1:1,000, Cell Signaling Technology, Danvers, MA) or total actin (1:2500, Chemicon), followed by monoclonal anti-mouse or polyclonal anti-rabbit horseradish peroxidase-conjugated secondary antibody, and visualized with an enhanced chemiluminescence system (Pierce, Rockford, IL), as previously reported [44, 46].

Groups of 3–5 P7 KO and WT pups were injected with PBS and 0.01 N HCl vehicle or IGF1 (10 u g/g, Cell Sciences) subcutaneously (sc) at time zero, followed at 6 h by ³H-dT (5 u Ci/g, sc), and were killed at 8 h. Cerebella were cleaned of meninges, removed from the brainstem, weighed in pre-tared microcentrifuge tubes and homogenized in dH₂O (volume = mass × 5). An aliquot was removed for determination of total isotope uptake into the tissue. In an equal aliquot, DNA was precipitated with 10% trichloroacetic acid, sedimented by centrifugation, and washed by resuspension and resedimentation. The final pellet was dissolved and counted along with the original aliquot in a scintillation spectrophotometer. ³H-dT incorporation was defined as the fraction of total radionuclide taken up by the whole cerebellum that was present in the precipitated DNA and was reported as percent incorporation [47, 48].

Two groups of three P7 KO and WT mice were injected sc with BrdU (50 u g/g) and killed at 2 h. Whole brains were removed and drop-fixed in fresh, cold 4% paraformaldehyde. Brains were serially dehydrated in 50, 60, 70, 80, 95 and 100% ethanol, 100% butanol and 100% xylene under vacuum and embedded in paraffin wax. Brains were then serially sectioned at 5 μm parasagittally, and sections were mounted on glass slides, with 5–6 sections/slide collected and numbered consecutively. Every fifth slide from WT and every fourth from KO (due to the smaller cerebellar size) plus two slides from either side of the central-most slide representing the central vermis were deparaffinized and rehydrated for anti-BrdU immunohistochemical processing. BrdU incorporated in DNA was detected by monoclonal antibody (Becton-Dickinson), amplified by biotinylated secondary antibody (Vectastain ABC kits), visualized by peroxidase reaction with DAB and subsequently counterstained with basic fuchsin. BrdU immunoreactive GNPs in the dorsal outer external germinal layer of lobules IV and VIII were counted by camera-lucida microscopy, and the BrdU labeling index quantified as the number of BrdU-positive cells/total cells × 100%. Four to eight sections from the hemispheres and four to eight sections from the vermis were counted for each pup [37, 48].

Data are expressed as mean ± SEM. Statistical comparisons were made by unpaired two-tailed Student’s t test or two-way ANOVA using Excel (Microsoft) or Vassarstats Website for Statistical Computation (http://​www.​vassarstats.​net), respectively.

Cerebellar GNPs increase expression of En2 postnatally in the inner EGL, concurrent with their cell cycle exit and early differentiation; therefore, En2 may be an important regulator of these processes. Using thymidine analog bromodeoxyuridine (BrdU) to label cells in S-phase, as previously published [48], we defined the labeling index [BrdU LI; (BrdU+/total cells) × 100%] of GNPs on postnatal day 7 (P7) of En2 KO and WT cerebellum (Figure 1A,B). BrdU-labeled GNPs were localized normally in the outer EGL in both genotypes. The BrdU LI was increased by 15% in En2 KO compared to the WT (Figure 1C), suggesting En2 may regulate precursor proliferation. Furthermore, genotype-dependent differences were even greater when proliferation was assessed specifically in the centrally localized vermis, a region that may express higher En2 levels [15, 24]: the BrdU LI was 23% in the WT, whereas it was increased to 30% in the KO (Figure 1D). These results are consistent with a model in which En2 normally suppresses S phase entry of GNP in the EGL.

Differences in GNP proliferation in vivo may reflect the effects of distinct environmental signals such as growth factors derived from Purkinje neurons, which are reduced in the mutants, or alternatively, action of cell autonomous signals, such as En2 expression [49]. Previously, we reported that in the absence of mitogens, specifically Shh, mouse GNPs in culture rapidly exit the cell cycle [38]. Therefore, WT and KO GNPs were isolated from the cerebellum and cultured at low density in defined media without growth factors, and DNA synthesis was assessed at 24 h by BrdU immunocytochemistry. The proportion of GNPs in mitotic S-phase was two-fold higher in En2 KO cells (Figure 1E,F). Further, there were similar numbers of live KO and WT GNPs counted at 24 h (WT = 143 ± 10.4; KO = 120 ± 9.8, total cells/five fields ± SEM; N = 5-6 per genotype; p = 0.14). These in vitro results recapitulate the in vivo finding that En2 KO GNPs fail to exit the cell cycle at the same rates as WT GNPs, an effect not likely due to changes in survival. These data also suggest that En2 participates in cell-autonomous regulation of the GNP cell cycle, independent of extracellular growth factors in the EGL. However, this raises the question, does En2 also function to modulate the effects of extracellular signals on cell cycle regulation?

Growth factors secreted by underlying Purkinje cells have been well characterized as regulators of GNP proliferation, differentiation and survival [49‐52]; therefore, abnormal GNP responses to growth factor signaling in the absence of En2 expression may contribute to the KO phenotype. To address this issue, we isolated En2 KO and WT GNPs and compared proliferation, differentiation and survival in culture without and with developmentally relevant extracellular growth factors.

Mitogenic growth factors, those that promote G1/S-phase transition, were added to defined media, and DNA synthesis in KO and WT GNPs was assessed by measuring ³H-deoxythymidine (³H-dT) incorporation. Sonic hedgehog (Shh), arguably the most well-recognized mitogen in postnatal cerebellar development [38, 50, 53‐55], elicited identical increases in DNA synthesis in both genotypes (Figure 2A). However, insulin-like growth factor 1 (IGF1) and high-dose insulin, known to act through the IGFI receptor [56], elicited two-fold greater increases in DNA synthesis of KO GNPs (Figure 2A). This differential genotype response was apparent at physiologic concentrations of IGF1 and above (Figure 2B). Since the dose-response profiles to IGF1 were similar between genotypes, the differential mitogenic responses were not a function of the specific dose employed, and the receptor-ligand binding kinetics are apparently unchanged. Furthermore, Shh and IGF1 together increased DNA synthesis synergistically; however, the differential genotype-specific responses persisted (Figure 2A). This suggests IGF1 may specifically stimulate enhanced proliferation in the absence of En2. To further explore the specificity of IGF1 effects, anti-mitogenic stimuli, fibroblast growth factor-2 (FGF2) and pituitary adenylate cyclase activating peptide (PACAP), which block cell cycle progression in mouse GNPs [38], were added to defined media, and DNA synthesis was assessed (Figure 2C). FGF2 treatment reduced WT GNP incorporation by 50%, whereas PACAP elicited a 25% reduction, and both FGF2 and PACAP together reduced incorporation by 65%; virtually identical changes in DNA synthesis were observed in KO GNPs. Moreover, these anti-mitogenic signals attenuated IGF1-stimulated increases in DNA synthesis in both genotypes. Nonetheless, IGF1 stimulated greater DNA synthesis in KO GNPs even in the presence of these anti-mitogenic stimuli (Figure 2C). No other growth factors we tested elicited differential genotype-specific responses in DNA synthesis, including epidermal growth factor (EGF), brain-derived neurotrophic factor (BDNF) and vascular endothelial growth factor (VEGF), nor did Wnt-3a-conditioned media (data not shown). These data suggest IGF1 signaling may be differentially regulated in the absence of En2 in cultured GNPs.

To determine whether IGF1 signaling is differentially regulated by the absence of En2 in vivo, we administered a single subcutaneous injection of IGF1 (10u g/gbw) to P7 KO and WT mice and measured total cerebellar tissue DNA synthesis using radiolabeled ³H-dT thymidine. As previously described for effects of peripherally injected bFGF on neonatal cerebellum [48], changes in the magnitude of ³H-dT thymidine incorporation into whole cerebellar homogenates (see [37], Figure one) closely paralleled changes in the labeling index of isolated GNPs (see [37], Figure four), supporting the value of this approach. With regard to IGF1, previous studies indicate that peripherally administered growth factor actively passes the blood-brain-barrier via a saturable transport mechanism and penetrates the brain parenchyma [41, 57]. Preliminary studies conducted at 4, 8 and 12 h post-IGF1 injection, with a subcutaneous ³H-dT (5u Ci/g) injection 2 h prior to sacrifice, indicated the greatest increase in DNA synthesis was observed at 8 h. As we detected in culture, IGF1 significantly increased total cerebellar DNA synthesis in both genotypes in vivo. However, in the absence of En2, IGF1 elicited a far greater increase in cerebellar DNA synthesis, increasing ³H-dT incorporation by 28% over saline injection in the KO, whereas the increase was only 9.7% in the WT (Figure 2D). These data recapitulate in vivo that IGF1-induced mitogenesis is differentially regulated in the absence of En2 and suggest En2 negatively regulates IGF1 signaling during postnatal cerebellar development.

Postnatal En2 expression begins in postmitotic GNPs in the inner EGL at, or just before, the start of differentiation and migration [15, 24], suggesting this gene may regulate these processes. However, En2 KO mice develop a mature, albeit smaller, IGL than the WT, suggesting En2 may not be required for these processes but rather may modulate them. To address GNP differentiation in the absence of En2, KO and WT cells were cultured in defined media for 24 h without and with growth factors, and differentiation was assessed by quantifying neurite-bearing cells, as previously described [58] (Figure 3A). In media lacking growth factors, an identical percentage of WT and KO GNPs extended neurites (WT = 16.5% ± 1.88; KO = 16.8% ± 0.74; percent bearing neurites ± SEM; p = 0.86). When cultured with well-characterized neuritogenic factors PACAP or IGF1 [51, 52], WT cells exhibited increased neurite outgrowth of 30% and 56%, respectively, over that occurring in defined media alone. However, En2 KO GNPs failed to respond to PACAP, whereas IGF1 only increased KO neurite outgrowth by 23%, a 59% reduction (p ≤ 0.05) in IGF1-stimulated outgrowth compared to WT levels (Figure 3B). Thus, in the absence of En2, GNPs exhibit reduced growth factor-induced neurite outgrowth, suggesting that En2 expression normally enhances differentiation. Interestingly, simultaneous administration of IGF1 and PACAP together “rescued” the KO GNP differentiation deficit, increasing neurite outgrowth to WT GNP levels (Figure 3B). These results may suggest there are convergent signaling pathways downstream of PACAP and IGF1 that together can promote GNP differentiation even in the absence of En2.

However, there may be an alternative model for the ability of PACAP plus IGF1 to overcome the neurite outgrowth deficiency observed in KO GNPs. Specifically, the anti-mitogenic effect of PACAP on both WT and KO GNPs, as above (Figure 2C), may allow cells to differentiate that would have otherwise continued to proliferate. Thus, any anti-mitogenic signal would be expected to promote IGF1-induced neuritogenesis. To address this issue, KO and WT GNPs were cultured with FGF2, another anti-mitogenic signal, and effects of IGF1 were examined. FGF2, a strong anti-mitogenic signal in both WT and KO GNPs (Figure 2C), elicited reductions in neurite outgrowth in both genotypes, suggesting it is not a pro-differentiation signal either (Figure 3C). FGF2 significantly attenuated IGF1-induced neurite outgrowth by more than 30% in both genotypes (Figure 3C). Thus, FGF2 represents both an anti-mitogenic and anti-neuritogenic signal that is equally active in both KO and WT GNPs. These findings are consistent with at least one report suggesting FGF2 inhibits neurite outgrowth in cerebellar GNPs [59]. Further, these data suggest KO GNP deficits in neurite outgrowth are not due to failed cell cycle exit, but rather a specific failure or delay in differentiation in the absence of En2.

In addition to regulating the cell cycle and differentiation, both PACAP and IGF1 exert trophic support for developing GNPs in the EGL [38, 52]. Therefore, differences observed in DNA synthesis and neurite outgrowth between WT and En2 KO GNPs could reflect differences in survival. To examine survival, cells were plated at low density (10,000 cells/cm²) to limit cell-cell contact and counted at 24 h in defined media without and with PACAP, IGF1, or PACAP and IGF1 together. Initial counts were taken at 2 h post plating to ensure equal numbers of cells were aliquoted and attached between genotypes. As reported previously, both PACAP and IGF1 increased the number of WT cells, and they exhibited a trend toward additivity when co-administered (Figure 3D). Significantly, En2 KO cells responded identically to WT GNPs, exhibiting similar increases in cell numbers when incubated with PACAP, IGF1 and combined factors (Figure 3D). Thus, the absence of En2 does not appear to be deleterious to GNP survival or to alter the trophic responses to growth factors. Further, these data suggest differences in DNA synthesis and neurite outgrowth observed between KO and WT GNPs are due to differentially regulated mechanisms underlying mitosis and differentiation, and not differences in cell survival or cell death.

As shown above, IGF1 is a pleiotropic signal in the developing cerebellum, regulating proliferation, neurite outgrowth and survival via its tyrosine-kinase receptor, the IGF1 receptor (IGF1R), which is activated through receptor autophosphorylation of tyrosine residues [60]. Activated IGF1R in turn binds and activates one of several phosphorylation cascades engaging different signal transduction pathways, with phosphotidylinositol-3-kinase (PI3K) being one of the most commonly activated in cerebellar GNPs [61‐63]. PI3K can induce activation of phosphokinase B/Akt (Akt), leading to inhibition of glycogen synthase-kinase-3-beta (GSK3beta), and can promote proliferation and survival [64, 65]. Alternatively, IGF1 signaling can activate ras/raf-MEK, which activates the mitogen-activated protein kinase (MAPK)/extracellular signal-regulated protein kinase 1 and 2 (ERK1/2), leading to survival and differentiation [66]. To determine which cascades may underlie genotype-specific effects of IGF1, protein levels and the activation ratios of phospho-Akt/Akt (P-Akt) and phopho-ERK1/2/ERK1/2 (P-ERK) were measured by Western blot. Initial comparisons were made between untreated, whole cerebellar lysates from P7 KO and WT mouse pups. There were no genotype differences in either P-Akt or P-ERK, suggesting that in the absence of En2, baseline pathway activation was similar to WT in developing cerebellum (Figure 4A).

To determine whether IGF1 signaling through PI3K-Akt or MAPK is altered specifically in GNPs in the absence of En2, cells were isolated from each genotype and cultured at cell densities similar to those used in the ³H-dT proliferation assays. To assess IGF1 activity, KO and WT GNPs were cultured in defined media for 2 h without IGF1 and were then pulsed for 30 min with vehicle or 10 ng/ml IGF1. Both KO and WT GNPs exhibited almost no P-Akt activation in vehicle-treated control cultures (Figure 4B). Conversely, IGF1 elicited identical, robust phosphorylation of Akt in both genotypes. Previous studies show that PI3K pathway activation, measured by phosphorylation of Akt, occurs within minutes of growth factor treatment and becomes maximal at 1 h [67]. However, our culture results on mitosis and differentiation were obtained following continuous growth factor exposure for 24 h. Therefore, experiments were repeated at several later time points, including IGF1 treatments of 15 min, 4 h and 8 h. Though IGF1 increased P-Akt early at 15 min, and to a lesser degree at 4 h and 8 h, there were no differences between the genotypes; further, phospho-GSK3-beta protein levels were also no different between genotypes (data not shown). Furthermore, IGF1 pulses did not significantly increase P-ERK at any time point, nor were P-ERK levels different between the genotypes (Figure 4C). This suggests that in our culture models, IGF1 preferentially activates the PI3K-Akt pathway, whereas the MAPK pathway is constitutively activated, similar to previous reports [68]. Genotype-specific differences in IGF1 responses do not appear to be mediated by these dominant upstream pathways.

Given the dominance of the PI3K pathway in our culture models, differences in En2 KO and WT GNP responses to IGF1 were explored further downstream of Akt. The downstream Akt target, mammalian target of rapamycin (mTOR), is known to regulate cell cycle progression as well as protein translation via activation of S6 kinase (S6K) and eukaryotic initiation factor 4B (eIF4B) [69]. In contrast to the upstream signals, IGF1 induced a differential genotype response in phospho-S6 kinase (P-S6K), eliciting a 25-fold increase in P-S6K levels in KO GNPs at 30 min, but inducing no significant increase in WT GNPs (Figure 4D). These data suggest that the differential genotype response of WT and KO GNPs to IGF1 may reflect a previously undefined interaction between En2 and downstream effector molecules of the IGF1-PI3K-Akt-mTOR signaling cascade.

Given our findings that the absence of En2 resulted in increased GNP proliferation and decreased differentiation, it follows that overexpression should elicit opposite effects: increased GNP cell cycle exit and differentiation. Previously, we demonstrated that ectopic En2 overexpression in embryonic cortical precursors altered neurogenesis [7], and additional studies in HEK293 cells confirmed that transfection produces overexpression of En2 cDNA (not shown). Using these same vectors, we overexpressed En2 by lipid transfection of P7 GNPs.

GNPs were isolated from P7 WT mice and transfected 2 h after plating with the En2-GFP cDNA vector or control GFP vector. Proliferation was assessed 24 h after the 5-h transfection period by a terminal 4-h pulse of BrdU, approximately 30 h post plating. While GNPs transfected with control GFP vector exhibited a BrdU LI of 6.54%, overexpression of En2 completely abolished BrdU labeling (Figure 5A-H; p ≤ 0.05; n = 8 per vector). In addition to cell cycle withdrawal, En2 overexpression elicited a 2.5-fold increase in WT GNPs exhibiting neuronal morphologies (i.e., neurite-bearing cells, Figure 5I, K) compared to GFP-transfected controls, indicating that En2 overexpression promotes cell cycle exit and early differentiation. Moreover, in GNPs from the KO mouse, En2 overexpression elicited effects that were nearly identical to WT cells, demonstrating a two-fold increase in cells exhibiting neuronal morphologies (Figure 5J,K). These data suggest the deficits in neurite outgrowth observed in KO GNPs (Figure 3) that are due to the developmental absence of En2 were rescued by acute re-expression of the transcription factor at P7. Thus, as predicted, En2 overexpression produced effects opposite to En2 deletion, reducing markers of proliferation while increasing markers of differentiation.

To further explore En2 effects, we performed additional studies in postnatal rats, which exhibit very similar cerebellar development [70, 71], because several antibodies were ineffective for murine cells, and postnatal rats were more readily available. En2 DNA and protein sequences are highly homologous across mammalian species, with ~96% homology between mouse and rat [72]. As observed in mouse, rat GNPs overexpressing En2 also exhibited a reduction in BrdU LI, by 67%, compared to GFP-only transfected cells (Figure 6A). However, changes in BrdU labeling may possibly reflect changes in cell cycle stage lengths. To more directly assess precursor status, we immunostained GNPs for proliferating cell nuclear antigen (PCNA). In contrast to BrdU, PCNA is expressed from S- through M-phase and has a long half-life, making it a more sensitive marker of the precursor cell compartment [73]. More than 30% of GNP exhibited PCNA immunoreactivity (Figure 6B), suggesting that many cells were proliferative precursors. Overexpression of En2 reduced PCNA immunostaining by 52% (Figure 6B), suggesting that gene overexpression induced cells to transition from precursors to postmitotic neurons. Additionally, similar to mouse GNPs, overexpression of En2 in rat GNPs resulted in a two-fold increase in cells exhibiting neuronal morphologies (Figure 6C). Furthermore, neurite-bearing cells characteristically express cytoskeletal proteins such as microtubule-associated protein 1b (Map1b), which is a known target for regulation by En[74]. En2 overexpression increased Map1b immunostaining in transfected GNPs by 27% (Figure 6D). Together, these data suggest En2 overexpression promotes rat GNP cell cycle exit similar to mice, suggesting En2 may have an evolutionarily conserved function in other mammalian species.

During postnatal cerebellar development, outer EGL GNPs proliferate robustly, producing a large population of internal granule layer neurons. Underlying Purkinje neurons regulate this process by producing extracellular growth factors, including Shh, IGF1 and PACAP [49, 50, 75]. However, cell intrinsic mechanisms also regulate the GNP transition from proliferation to post-mitotic differentiation [76]. While mechanisms remain undefined, we present evidence suggesting En2 serves to promote GNP cell cycle exit. These findings parallel evidence that outer EGL GNPs that are devoid of En2 remain proliferative while inner EGL GNPs begin expressing En2 as they exit the cycle [15, 24]. Indeed, when isolated in culture, GNPs from the WT mouse were less proliferative than those from the KO. These observations suggest En2 normally inhibits cell cycle progression in a cell-autonomous fashion, a model that is also supported by En2 overexpression that reduces S phase entry and precursor PCNA expression.

Purkinje neurons secrete growth factors, such as Shh, FGF, IGF1, BDNF and PACAP, that support local GNP proliferation, survival and differentiation [37, 46, 49, 52, 65, 77‐79]. We find En2 expression modulates GNP responses to IGF1, suggesting that the onset of En2 expression contributes to cell cycle exit and differentiation in the growth factor-rich environment of the EGL. In En2 KO GNPs, IGF1 induced two- to three-fold greater increases in DNA synthesis in culture and three-fold greater stimulation of cerebellar DNA synthesis in vivo. Importantly, En2 did not appear to regulate overall cell cycle machinery or signaling by other growth factors that employ signaling systems similar to IGF1, such as tyrosine-kinase engaging receptors or PI3K/Akt or MAPK pathways. Rather, in the absence of En2, IGF1 markedly upregulated P-S6K, an effect not observed in WT cells. While P-S6K is a well-defined regulator of mitogen-induced cell cycle progression [69], there were no genotype differences in upstream PI3K-Akt-GSK3beta or MEK/ERK pathways. To our knowledge, this is the first report of interactions between En2 expression and IGF1 signaling.

These initial observations require further study to (1) define relationships of P-S6K activation to enhanced mitogenesis and (2) identify mediating upstream regulators, potentially members of the mTOR complex 1 family [69, 80, 81]. Indeed, extracellular mitogens promote mTOR signaling, which acts through S6K and eIF4B to promote cell proliferation [80]. eIF4B phosphorylation by S6K in turn is growth-factor dependent and important for promoting ribosome-mRNA binding, and consequent growth and proliferation [69]. Significantly, Engrailed 2 protein has previously been shown to bind eIF4E [82]. Thus, Engrailed 2 may interact with other members of the eIF4 family, such as eIF4B, providing a mechanism through which En2 alters downstream mTOR signaling. This might explain the increases in IGF1-induced proliferation we observe both in vitro and in viv o. However, what underlies the increased IGF1-induced S6K phosphorylation in En2 KO cells? Independent of Akt-mTOR signaling, IGF1 activates through PI3K the 3-phosphoinositide-dependent protein kinase 1 (PDK1), which is known to phosphorylate S6K (Figure 7) [83]. Thus, PDK1 is a candidate in addition to mTOR for further study. Furthermore, since S6K itself negatively regulates IGF1 signaling by phosphorylating the insulin response substrate 1 (IRS1) [84], En2 may disrupt this inhibitory feedback loop, leading to increased S6K phosphorylation and activation via mTOR-dependent and -independent pathways downstream of PI3K. Thus, our data suggest En2 may function normally to attenuate mTOR-mediated proliferation in a growth factor-rich environment, though the mechanisms by which this may occur remain to be explored.

In addition to altered mitogenesis, En2 KO GNPs grew fewer neurites in response to neuritogenic factors, IGF1 and PACAP, though their combination synergistically overcame differentiation deficits. Decreased IGF1-induced neuritogenesis was not merely due to failed cell cycle exit, a possibility because the PACAP rescue of reduced neurite outgrowth was accompanied by inhibition of proliferation [38]. Rather, treatment with another anti-mitogen, FGF2, did not rescue the differentiation defect, but in fact attenuated IGF1-stimulated neuritogenesis in both genotypes. Significantly, GNP survival was also not compromised in the absence of En2, nor was growth factor-induced trophism. In aggregate, these data suggest that a complex array of mitogenic, trophic and differentiative signals within the EGL act in concert with cell-autonomous gene expression (i.e., En2 expression) to execute GNP maturation.

Previous evidence indicates that multiple interactions of intrinsic and extrinsic signals control GNP proliferation. With regard to extrinsic signals, we previously found that PACAP inhibits Shh-induced GNP mitosis by upregulating adenylate cyclase [38], whereas others show similar roles for bone morphogenetic proteins via Smad [53, 85]. Conversely, in the current study, we find IGF1 and Shh synergistically stimulate DNA synthesis. Furthermore, interactions between intrinsic signals and growth factors are also known. For example, transcription factor Atoh1/MATH1 expressed in EGL precursors promotes Shh-induced proliferation [86] by regulating expression of downstream target, Gli1 [87]. Similar promitogenic effects have been defined for Zic family members, including 1, 2 and 4, as well as ATF5 [88‐90]. On the other hand, ZNF238/RP58 is expressed as GNPs exit the cell cycle, and overexpression inhibits proliferation [91, 92], findings that parallel our observations with En2. Thus, a number of gene-growth factor interactions likely contribute to GNP development. That no other mitogenic or anti-mitogenic signal elicited genotypic differences in En2 KO and WT GNPs suggests there is a specific yet indirect interaction between En2 and IGF1 signaling (Figure 7), with one feature being altered S6K activation. The expression of En2 by GNPs may facilitate their transition from proliferation to differentiation in an environment rich in mitogenic signals [93], thereby allowing subpopulations of cells to continue cycling while others exit and differentiate. En2 expression has been reported to be greater in the vermis than the hemispheres [15, 24], which may be one mechanism explaining the delay between vermis and hemisphere GNP proliferation during development [23].

Previously, we found that ectopic En2 overexpression in embryonic cortical precursors increased proliferation [7]. This stimulatory effect, opposite to what we observe here, is consistent with the universal expression of En2 in prenatal hindbrain progenitors and suggests that before birth En2 serves to maintain proliferation. In the current study of postnatal GNPs, En2 overexpression completely abolished BrdU labeling in WT mouse cells and reduced markers of proliferation (PCNA and BrdU) in rat GNPs. Additionally, En2 overexpression more than doubled the proportion of mouse and rat GNPs that exhibited neurite outgrowth. Significantly, neurite outgrowth was stimulated even in KO GNPs, suggesting that acute En2 expression may reverse morphological deficits associated with the absence of En2 for all of development, though additional studies are warranted. Furthermore, despite biological differences in GNPs from mice and rats [38], En2 overexpression produced similar phenotypes in both species, suggesting En2 function is evolutionarily conserved. The opposite consequences of En2 deletion and overexpression on GNP maturation suggest En2 promotes cell cycle exit and differentiation during postnatal cerebellar development.

One limitation to our current model, however, remains the unexplored role of En2 in prenatal cerebellar neurogenesis. While postnatal granule neurogenesis ultimately dictates final cerebellar morphology and size, prenatal patterning events organize the developing cerebellum and delineate neuronal progenitor cell populations, ultimately producing a cytoarchitectural framework for future axonal pathway elaboration [15, 94‐97]. In the early cerebellar anlagen, En2 expression is ubiquitous, as is progenitor cell proliferation; therefore, En2 is unlikely to promote cell cycle exit and differentiation at this time. Previous studies suggest En2 may function during this period to specify numbers as well as types of progenitor cells that will give rise to various cerebellar neuronal populations [14, 16, 50, 95, 96, 98]. Thus, our characterization of postnatal En2 function is likely a time-locked developmental phenomenon that cannot address its prenatal activities. Rather, our demonstration of increased GNP proliferation may potentially suggest why En2 KO mice are capable of producing cerebella at all, despite the prenatal insult.

We reported previously that intronic polymorphisms (A-C haplotype) of human EN2 are associated with ASD [7, 9] and that the presence of these SNPs results in altered transcription factor binding as well as increased levels of gene expression [99]. Furthermore, our recent studies indicate that in human cerebellum, the disease-associated allele produces increased EN2 expression [100]. If the En2 activity defined here in rodent studies is relevant to primates, how then would altered EN2 expression be expected to affect humans? Increased EN2 expression during postnatal cerebellar development, the period when the majority of human granule neurons are generated, would likely produce a granule neuron deficit by eliciting premature cell cycle exit. However, as with the mice, such a simple prediction is unwarranted because the prenatal effects of gene alteration remain undefined. But at the broader level, children with ASD exhibit a range of cerebellar structural abnormalities, including a diminished vermis as well as enlarged hemispheres in which granule neuron numbers are dysregulated [3, 101‐103], a phenotype to which EN2 may contribute. Indeed, in mouse models, both gene deletion as well as ectopic overexpression produces cerebellar hypoplasia [13, 18, 20]. In addition, cerebellar abnormalities have been described in human and mouse studies of autistic phenotypes in tuberous sclerosis as well as a number of neuropsychiatric disorders including schizophrenia, attention deficit hyperactivity disorder, and cognitive and language disabilities [35, 104‐106].

Another interesting implication of our studies is identification of an interaction between En2 expression and IGF1, the latter being an important pleiotropic growth factor associated with disease as well as a possible therapeutic target. Riikonen et al. [107] reported reduced IGF1 in cerebrospinal fluid (CSF) of ASD children compared to age-matched controls and that CSF levels correlated with head circumference in ASD, but not control children. Given the En2-IGF1 interaction we describe, one might speculate whether altered EN2 expression may coincide with abnormal IGF1 to contribute to ASD pathogenesis, a question that remains to be investigated. From a therapeutic perspective, the Akt-mTOR-S6K pathway is dysregulated in multiple animal models of monogenic causes of ASD including fragile X mental retardation [108], Rett syndrome [109] and tuberous sclerosis [110], whereas IGF1 ligands may improve neurodevelopmental symptoms in Rett [111] and the SHANK3 autism-related mouse model of Phelan-McDermid syndrome [112]. It is intriguing that yet another autism-associated gene, in this case EN2, implicates disordered Akt-mTOR-S6K signaling in the disease phenotype [113]. The exact molecular mechanisms mediating En2’s modulation of IGF1 signaling remain to be elucidated.