Divergent functions for eIF4E and S6 kinase by Sonic hedgehog mitogenic signaling in the developing cerebellum

SUMMARY

Cerebellar development entails rapid peri-natal proliferation of cerebellar granule neuron precursors (CGNPs), proposed cells-of-origin for certain medulloblastomas. CGNPs require insulin-like growth factor (IGF) for survival and Sonic hedgehog (Shh)--implicated in medulloblastoma--for proliferation. The IGF-responsive kinase mammalian Target of Rapamycin (mTOR) drives proliferation-associated protein synthesis. We asked whether Shh signaling regulates mTOR targets to promote CGNP proliferation despite constitutive IGF signaling under proliferative and differentiation-promoting conditions. Surprisingly, Shh promoted eIF4E expression, but inhibited S6 kinase (S6K). In vivo, S6K activity specifically marked the CGNP population transitioning from proliferation-competent to post-mitotic. Indeed, eIF4E was required for CGNP proliferation, while S6K activation drove cell cycle exit. PP2A inhibition rescued S6K activity. Moreover, Shh up-regulated the PP2A B56γ subunit, which targets S6K for inactivation and was required for CGNP proliferation. These findings reveal unique developmental functions for eIF4E and S6 kinase wherein their activity is specifically uncoupled by mitogenic Shh signaling.

INTRODUCTION

The kinase mammalian target of rapamycin (mTOR) integrates extracellular and intracellular signals to regulate cell growth in mammalian cells. Growth factors, nutrients and adenosine triphosphate (ATP) levels converge upon mTOR to propagate signals that promote translation of proliferation, metabolism, and growth-related messenger RNAs (mRNA), enabling cell cycle progression and cell size increase (Hay and Sonenberg, 2004; Ma and Blenis, 2009). Studies in cell lines have established that activated mTOR phosphorylates its substrates, 4E-binding proteins (4E-BP) and ribosomal protein (rp) S6 kinase (S6K), in parallel to promote simultaneous (a) release of eukaryotic initiation factor 4E (eIF4E) from inhibition and (b) activation of rpS6, thus enhancing activity of downstream effectors to promote synthesis of proteins required for cell cycle progression (Gingras et al., 2001). However, in the case of primary cells and in vivo systems, it is unclear how the mRNA translation machinery responds to environmental cues regulating proliferation or differentiation, an energetically expensive process that frequently entails cell type-specific morphological adaptations. This issue is particularly highlighted in the central nervous system (CNS), wherein cell numbers are rigorously controlled and mature neurons achieve a volume that is much larger than that of dividing progenitor cells, due to extension of axonal and dendritic processes or myelin sheath production as is the case in glial cells (Altman J. and Bayer, 1997).

The developing mammalian cerebellum is an ideal system for investigating mTOR pathway regulation in proliferating neural progenitors. The cerebellum is characterized by a small number of cell types, which undergo well-characterized developmental programs to achieve the final stereotypical architecture of the mature cerebellum, enabling this brain structure to process cues that regulate coordination, movement, and aspects of motor learning (Altman J. and Bayer, 1997). Moreover, aberrant proliferation of cerebellar neural precursors is implicated in medulloblastoma, the most common solid malignancy of childhood (Wechsler-Reya and Scott, 2001). Thus, studies of cerebellar development yield insight into normal and tumorigenic CNS growth control mechanisms.

Cerebellar granule neuron precursors (CGNPs) undergo a burst of rapid, post-natal expansion in the external granule layer (EGL) on the dorsal surface of the cerebellum (Hatten, 1998). This expansion phase lasts approximately 2 weeks in mice and 12 months in human infants. When CGNPs exit the cell cycle, they migrate internally and differentiate into interneurons, where they function as signal integrators between mossy fibers and Purkinje neurons (Altman J. and Bayer, 1997). In contrast to its roles in neural cell fate specification or differentiation elsewhere in the nervous system (Dutton et al., 1999a; Dutton et al., 1999b; Marti et al., 1995), Sonic hedgehog (Shh) serves as a mitogen for CGNPs (Dahmane and Ruiz-i-Altaba, 1999; Wallace, 1999; Wechsler-Reya and Scott, 1999). Shh binds to its receptor Patched (Ptc), lifting inhibition of the signaling pathway activated by the transmembrane protein Smoothened (Smo) (Ho and Scott, 2002), and resulting in activation of Smo target genes, including Ptc itself (Hepker et al., 1997). Downstream effectors of Smo that regulate CGNP proliferation include Gli transcription factors, N-myc, IRS1, YAP1, and the microRNA cluster miR17/92 (Dahmane et al., 2001; Fernandez et al., 2009; Kenney et al., 2003; Northcott et al., 2009; Oliver et al., 2003; Parathath et al., 2008; Uziel et al., 2009).

In addition to requiring Shh for proliferation, CGNPs also require signaling by insulin-like growth factor (IGF) for cell survival (D'Mello et al., 1997; Dudek et al., 1997). Notably, aberrant activation of the Shh and IGF signaling pathways is implicated in medulloblastoma (Hahn et al., 2000; Hartmann et al., 2005; Wechsler-Reya and Scott, 2001). IGF positively regulates the mTOR pathway (Corradetti and Guan, 2006) and CGNPs are continuously exposed to IGF in vivo and in vitro (D'Mello et al., 1997; Ye et al., 1996). Previous reports demonstrated that in addition to providing pro-survival signals, IGF signaling cooperates with the mitogenic Shh pathway to promote stabilization of N-myc, a Shh target required for CGNP proliferation and up-regulated in Shh-associated medulloblastoma (Kenney et al., 2004; Northcott et al., 2009; Pomeroy et al., 2002). However, CGNPs only proliferate in response to Shh (Wechsler-Reya and Scott, 1999) despite constitutive IGF pathway activation. This conundrum suggests that IGF-independent mechanisms exist downstream of Shh to promote preferential activity of mTOR substrates during the CGNP proliferation phase.

To test this hypothesis, we examined how Shh signaling affects expression and activity of mTOR pathway components in CGNPs. Surprisingly, we observed differential regulation of eIF4E and S6K activity, two of the best-characterized mTOR effectors, in response to Shh in proliferation-competent CGNPs. Our findings contrast with models that have been based on serum-stimulated cell line studies, which have indicated that eIF4E and S6K function cooperatively to promote proliferation (Burnett et al., 1998; Corradetti and Guan, 2006; Fingar et al., 2004; von Manteuffel et al., 1996). In primary CGNP cultures we correlated eIF4E expression with Shh-induced proliferation, while S6K activity was associated with CGNP cell cycle exit; indeed phosphorylation of the S6K substrate rpS6 marked the population of cells transitioning from the proliferative to the post-mitotic state in the neonatal cerebellum in vivo. The positive effects of Shh on eIF4E were at the transcriptional and protein up-regulation level, while the suppression of S6K activity involved protein phosphatase 2A (PP2A) activity, specifically through Shh-mediated up-regulation of the B56γ subunit, which targets PP2A to dephosphorylate and inactivate S6K.

RESULTS

Proliferating CGNPs differentially regulate mTOR substrates

To address the role of mTOR signaling in cerebellar development we used a well-characterized primary culture system of cerebellar granule neurons wherein isolated CGNPs from post-natal day (PN) 5 mice are maintained under serum-free conditions, in the presence of IGF which promotes their survival and cooperates with Shh to maintain proliferation (Dudek et al., 1997; Kenney et al., 2004). Continued proliferation is promoted by addition of Shh protein (Shh-treated) to the culture medium; cell cycle exit and differentiation into mature neurons occurs in the absence of Shh (vehicle-treated) (Kenney and Rowitch, 2000; Wechsler-Reya and Scott, 1999). When we examined levels and activity of mTOR pathway components in vehicle- and Shh-treated CGNPs by western blot analysis and densitometry, we found that proliferating (Shh-treated) CGNPs show up-regulation of proteins consistent with a need for increased cap-dependent mRNA translation, specifically mTOR itself, the mTOR partner Raptor, eIF4G, and other eIF4F components (Figure 1A, S1A). We also observed that Shh treatment promoted increases in total eIF4E and 4EBP2 levels, as well as 4EBP2 phosphorylation, suggesting that the negative regulation 4EBP2 has on eIF4E is relieved.

Figure. 1.

Shh-treated proliferating CGNPs show increased levels of translation initiation components and suppressed S6 kinase activity. (A) Western blot analysis of protein isolated from vehicle- or Shh-treated CGNP primary cultures. Shh treatment stimulates mTOR activity as demonstrated by eIF4G and 4EBP2 hyperphosphorylation but blocks S6 kinase activity measured by rpS6 phosphorylation. Levels of the mTOR effector eIF4E increase in Shh-treated CGNPs. (B) Quantification of western blots by densitometry. Phosphorylated protein levels were compared to total protein levels after normalization to β-Tubulin n=3, (**p<0.002, ***p<0.0005). (C to I) Immunofluorescence analysis of vehicle- and Shh-treated CGNPs plated on coverslips. eIF4E increases in Shh-treated CGNPs (D compared to C) while phospho-S235/236-rpS6 accumulates in vehicle-treated CGNPs (E compared to F). Total rpS6 levels do not change in proliferating or differentiated cells (G, H). Furthermore phospho-S235/236-rpS6 is expressed in vehicle-treated CGNPs marked by the differentiation marker, MEF2D (I arrowheads), while eIF4E protein colocalizes with PCNA positive cells (J arrowheads) in Shh-treated CGNPs. (K) Quantification of immunostaining in C-H, n=3, (***p<0.0001, **p<0.001). (L) Percentage of cells expressing eIF4E only, PCNA only or both eIF4E and PCNA in Shh-treated CGNPs. (M) Percentage of cells expressing MEF2D only, phospho-rps6 only or both MEF2D and phospho-rpS6 in vehicle-treated CGNPs.

Surprisingly, Shh suppressed S6K activity as determined by rpS6 phosphorylation (Figure 1A, S1B). When we compared the levels of phosphorylated to total protein we found that in both vehicle- and Shh-treated CGNPs 4EBP2 is phosphorlyated; however the amount of total rpS6 that is phosphorylated at either Serine residue in Shh-treated CGNPs is significantly decreased (Figure 1B). Previous reports from studies conducted largely in cell lines have indicated that mTOR phosphorylates both 4E-BP and S6K (Fingar et al., 2004; von Manteuffel et al., 1997) in parallel but our results suggest that mechanisms exist to preferentially specify phosphorylation of these mTOR substrates in primary neural precursors responding to mitogenic Shh signaling.

To further demonstrate that 4EBP2 phosphorylation combined with Shh-driven increases in total eIF4E levels results in active eIF4E we performed immunoprecipitation experiments to pull down eIF4E with eIF4G, the scaffold protein required for formation of the translation initiation complex. eIF4E only binds to eIF4G when it has been released from the negative inhibition by 4EBPs (Gingras et al., 1999; Pause et al., 1994) . As shown in Figure S1C eIF4E-eIF4G complexes form in both vehicle- and Shh-treated CGNPs, but at much higher levels in the presence of Shh .

To further investigate a role for eIF4E downstream of Shh in CGNP proliferation we stained primary CGNP cultures with antibodies against eIF4E and phospho-rpS6. Consistent with our western blot results, eIF4E protein levels significantly increased in Shh-treated CGNPs (Figure 1C, 1D, 1K) while phospho-rpS6 decreased (Figure 1E, 1F, 1K). As a control we also stained for total rpS6 in vehicle- and Shh-treated CGNPs and found no significant differences (Figure 1G, 1H, 1K). We also analyzed PCNA in CGNPs as an indicator of proliferation, and levels of MEF2D, a transcription factor induced during CGNP differentiation (Fogarty et al., 2007), to address whether eIF4E expression is linked to proliferation and S6K activity with differentiation on a per cell basis. As shown in Figure 1J, PCNA positive cells (green) also express eIF4E (red). Quantification of the number of single and double positive cells further demonstrates that eIF4E correlates with CGNP proliferation (Figure 1L). Furthermore, the majority phospho-rpS6 positive cells (red) also co-label with MEF2D (green) protein (Figure 1I, M), indicating an association between S6 kinase activity and CGNP cell cycle exit.

To determine whether our in vitro observations linking eIF4E with proliferation in CGNPs recapitulated the in vivo protein accumulation pattern, we carried out immunohistochemistry for eIF4E in the mouse cerebellum at PN 7 (Hematoxylin and eosin-stained sagittal section shown in Figure 2A). As shown in Figure 2B, eIF4E was strongly expressed in the EGLa, the region wherein CGNP expansion takes place, as well as Purkinje neurons. Cells bearing eIF4E largely overlapped with PCNA-positive cells (Figure 2C, C’) in the EGLa (the mitotic region of the EGL). In contrast, as shown in Figures 2D and 2E, eIF4E was excluded from cells of the EGLb (the post-mitotic region of the EGL) that express the cyclin dependent kinase (cdk) inhibitor p27 or NeuN, markers of CGNP cell cycle exit. eIF4E immunostaining did not reflect astrocytic expression, as astrocyte processes stained positive for glial fibrillary acidic protein (GFAP) (Figure 2F, F’), but were not eIF4E-positive.

Figure. 2.

Evaluation of eIF4E and phospho-S235/236 rpS6 localization in vivo. (A) Hematoxylin and eosin staining of wild type (wt) PN7 sagittal cerebellar sections. eIF4E is elevated in CGNPs located in the EGL (B). eIF4E (green) also colocalizes with the proliferation marker PCNA (red) (C). Higher magnification of area indicated in C further illustrates that cells in the EGL label with both eIF4E which is cytoplasmic and nuclear PCNA, arrows C’. eIF4E is not present in p27 (D) or NeuN (E) positive cells, which indicate post-mitotic or differentiated neurons, respectively. GFAP staining marks astrocyte cell bodies and processes that extend into the EGL (F). Confocal microscopy on area indicated in F demonstrates that eIF4E positive cells are on a separate plane from the astrocyte processes (arrows) (F’). (G) H&E staining on inner folia of wt PN7 sections to define area used to analyze phospho-S235/236-rpS6 protein, an indicator of S6 kinase activity. (H) Cells with increased S6 kinase activity are restricted to specific regions within EGL, indicated by arrows. Phospho-S235/236-rpS6 does not colocalize with p27 (I), NeuN (J and J’), or the astrocyte marker GFAP (K). Confocal microscopy on area defined in K confirms that astrocyte processes run along the cell membrane of phospho-S235/236-S6 positive cells in the EGL (arrows) (K’). Scale bars=100μm.

We also analyzed phospho-S6 levels in vivo. We observed a unique pattern of rpS6 phosphorylation in the EGL wherein phospho-rpS6 positive cells marked a specific cell population bordering the EGLa and EGLb (Figure 2H), where CGNPs transition from proliferating to post-mitotic. Additionally, Purkinje neurons and cells of the IGL were also positive. Interestingly phospho-rpS6 staining does not co-localize with previously described markers of post-mitotic (p27, Figure 2I) or differentiated granule cells (NeuN, Figure 2J, J’) in the EGL, indicating that these phospho-prS6-positive cells have not completed cell cycle exit. The absence of phospho-rpS6 staining in GFAP-positive cells suggests that in the EGL S6K activity may play a novel CGNP-specific role in mediating the transition from cycling to a non-cycling state (Figure 2K, K’).

Shh effects on eIF4E do not require mTOR activity

It has been reported that Shh signaling induces rapid up-regulation of cell proliferation-related mRNAs through either direct or indirect mechanisms (Katoh and Katoh, 2009). To test whether eIF4E is a transcriptional target of Shh signaling we isolated RNA from vehicle and Shh-treated CGNPs at different time points and then measured eIF4E transcripts by quantitative PCR (qPCR). As shown in Figure 3A, after 24 hours of Shh treatment a significant two-fold induction in eIF4E transcripts was observed (p=0.0005). The delay and levels of eIF4E induction compared to Gli1 (data not shown) suggest that Shh up-regulates eIF4E through an indirect mechanism. The induction of eIF4E was prevented when Shh-treated CGNPs were incubated in the presence of the Smoothened antagonist, SANT-2 (Figure 3A). To determine whether eIF4E up-regulation is independent of mTOR activity we incubated Shh-treated CGNPs with the mTORC1 inhibitor rapamycin for increasing periods of time. We then isolated RNA or protein and looked at effects on eIF4E and proliferation. As shown in Figure 3A, rapamycin did not significantly reduce eIF4E expression. Exposure to rapamycin for increasing periods of time further suppressed rpS6 phosphorylation and blocked 4E-BP2 phosphorylation, indicating that rapamycin is inhibiting mTORC1 kinase activity in CGNPs (Figure 3B). Rapamycin did not decrease eIF4E or cyclin D2 protein levels at 24 hours (Figure 3B), indicating that rapamycin-sensitive mTOR is not required for their maintenance. Interestingly, 24 hours of rapamycin treatment did not impair CGNP proliferation as determined by quantification of PCNA+ cells. Decreases in CGNP proliferation were only observed after 48 hours of rapamycin treatment (Figure 3D), by which time rapamycin may also be impacting upon mTOR:Rictor and impairing Akt activity through inhibition of S473 phosphorylation (Sarbassov et al., 2006). These results suggest that the increases in eIF4E are likely to be mediated through a Shh-dependent mechanism, independent of mTOR.

Figure. 3.

Shh induction of eIF4E does not depend on rapamycin-sensitive mTOR activity. (A) qPCR analysis performed on CGNPs shows that eIF4E expression increases over periods of prolonged Shh incubation suggesting an indirect up-regulation of eIF4E mRNA n=3; (***p=0.0005). Incubation with mTORC1 inhibitor rapamycin (10nM) for 12 or 24 hours did not effect eIF4E induction, however the Smoothened antagonist, SANT-2 (100nM), prevented Shh-mediated eIF4E induction (**p=0.0012), (B) Western blot analysis from CGNPs treated with the mTORC1 inhibitor, rapamycin (10nM), for increasing periods of time. Suppression of mTORC1 activity, confirmed by 4EBP2 and rps6 phosphorylation, did not alter eIF4E or cyclin D2 levels. (C) Western blot analysis from vehicle- and Shh-treated CGNPs treated with rapamycin (10nM) for 12 hours to confirm suppression of mTOR activity in untreated CGNPs based on rpS6 phosphorylation. (D) Quantification of proliferation differences based on PCNA staining in vehicle- and Shh-treated CGNPs after addition of rapamycin for either 24 or 48 hours n=4, (**p<0.001).

CGNPs require eIF4E for proliferation

We next wished to determine whether manipulation of eIF4E levels affects CGNP proliferation. We first carried out loss-of-function analysis using lentiviruses carrying short hairpin RNA sequences targeting eIF4E. We found that 3 out of the 5 shRNA lentiviruses tested in a murine cell line were effective in knocking down eIF4E (Figure S2) and then used those lentiviruses as a pool to infect CGNPs. Effects on eIF4E were specific as determined by comparison of eIF4E levels in CGNPs infected with control (GFP-targeting) lentiviruses and by assessment of proteins not expected to change by eIF4E knock-down. Although the infection efficiency was only approximately 20% in Shh-treated CGNPs (Figure S2), we observed decreased levels of cyclin D2, eIF4G, IRS1 in eIF4E-knocked-down CGNPs. We previously reported that insulin receptor substrate 1 (IRS1) is required for Shh-mediated proliferation (Parathath et al., 2008), thus these results hint at reduced CGNP proliferation in the absence of eIF4E. Interestingly, we also observed a recovery of rpS6 phosphorylation in these cells (Figure 4A), indicating rescue of S6K activity, known to de-stabilize IRS1 in CGNPs (Parathath et al., 2008). In contrast, phosphorylation of 4E-BP2 was not affected by eIF4E loss, in keeping with our observation (Figure 1A) that 4E-BP2 and S6K phosphorylation are independently regulated in Shh-treated CGNPs. These results indicate that an inverse relationship exists between eIF4E expression and S6K activity, in that when eIF4E is present, S6K activity is suppressed, but when eIF4E is lost and subsequently CGNP proliferation decreases, S6K activity returns.

Figure. 4.

eIF4E expression is required for the full mitogenic response of CGNPs to Shh. (A) Western blot analysis from CGNPs infected with a pool of lentiviruses containing sequences targeting eIF4E (see also Figure S1). Loss of eIF4E expression results in decreased proliferation, measured by cyclin D2, as well as recovery of S6 kinase activity. (B to D) Immunohistochemical analysis of CGNPs infected with shRNA containing lentiviruses. Effects on proliferation quantified by measuring BrdU incorporation in untreated (B), Shh-treated, GFP shRNA infected (C) and Shh-treated eIF4E shRNAs infected (D) CGNPs demonstrating a 50% reduction in cycling cells n=3; (*p<0.02) (E) when eIF4E was knocked down. (F) Western blot analysis from β-actin-Eif4e or WT CGNPs demonstrating that increased eIF4E levels are sufficient to promote Shh-independent proliferation. Vehicle-treated β-actin-Eif4e CGNPs have increased cyclin D2 levels compared to WT. (G) Immunohistochemical analysis of proliferation by BrdU incorporation (green) in vehicle and Shh-treated WT or β-actin-Eif4e CGNPs. (H) Differences in proliferation quantified by comparing number of BrdU incorporated cells in vehicle-treated WT and β-actin-Eif4e CGNPs or Shh-treated WT and β-actin-Eif4e CGNPs. n=3; (***p<0.0003).

Reduced cyclin D2 and IRS1 levels in eIF4E-knocked down CGNPs suggests that loss of eIF4E decreased proliferation of these cells (Figure 4A). To quantitatively determine the effect of eIF4E loss on proliferation in Shh-treated CGNPs, we measured BrdU incorporation in control-infected and eIF4E shRNA-infected cells following a 2-hour BrdU pulse (Figure 4B-D). eIF4E knockdown reduced BrdU incorporation by 50% compared to Shh-treated control-infected CGNPs, as determined by quantification of immunofluorescent staining (p<0.02) (Figure 4E); however eIF4E loss did not compromise CGNP survival, as we saw no increase in cleaved caspase 3-positive cells (data not shown). These results suggest a requirement for eIF4E in CGNP proliferation, as well as indicating a role for S6K activity as CGNPs leave the cell cycle.

eIF4E overexpression is sufficient to drive Shh-independent CGNP proliferation

Since we observed that eIF4E is up-regulated in the presence of Shh and is required for CGNP proliferation, we wished to determine how eIF4E over-expression affects CGNP proliferation. To this end we used a recently described in vivo mouse model, in which eIF4E is ubiquitously over-expressed under control of the β-actin promoter (β-actin-Eif4e) (Ruggero et al., 2004). As adults, these mice develop lymphomas, angiosarcomas, adenocarcinomas and hepatomas but they do not develop medulloblastomas, indicating that eIF4E over-expression alone is not sufficient to transform CGNPs in vivo. We isolated CGNPs from β-actin-Eif4e mice or WT age-matched controls mice and established primary cultures. As demonstrated by western blot analysis, vehicle-treated β-actin-Eif4e CGNPs have increased eIF4E levels compared to vehicle-treated WT but similar to Shh-treated WT and β-actin-Eif4e CGNPs (Figure 4F). Furthermore the increase in eIF4E is sufficient to drive CGNP proliferation independent of Shh as evidenced by cyclin D2 levels in vehicle-treated β-actin-Eif4e CGNPs. This ability of β-actin-Eif4e CGNPs to proliferate in the absence of Shh was confirmed by quantification of BrdU incorporated cells (Figure 4G, H), which demonstrates that increased eIF4E results in twice as many cycling cells compared to WT (p<0.0003). Interestingly though, increased eIF4E does not cooperate with Shh to promote proliferation as the number of BrdU positive cells in Shh-treated WT and β-actin-Eif4e CGNPs were similar. These results suggest that increased eIF4E levels alone can promote CGNP proliferation independent of Shh; however this increase is insufficient to cause transformation as β-actin-Eif4e CGNPs still maintain cell intrinsic cues to exit the cell cycle and initiate differentiation programs.

S6K drives CGNP cell cycle exit

Based on our initial western blot and immunofluorescence observations, S6K activity appears to be induced at a point when CGNPs are poised to exit the cell cycle, and pro-proliferative eIF4E activity antagonizes S6K activity. To further investigate the connection between S6K and cell cycle exit in CGNPs, we asked how CGNPs responded to differentiation cues with respect to S6K activity. We incubated Shh-treated CGNPs with bFGF, which has been reported to induce CGNP differentiation (Fogarty et al., 2007). Under conditions of bFGF treatment, a dramatic upregulation of phosphorylated rpS6 occurred (Figure 5A). Western blot analysis further supports the conclusion that FGF treatment caused cell cycle exit as indicated by decreased cyclin D2 and N-myc levels (Figure 5A). We also evaluated activity of other kinases and consistent with our previous studies (Kenney and Rowitch, 2000; Parathath et al., 2008) we observe no effects of Shh on the activity of these kinases (Figure 5A). Additionally, incubating Shh-treated CGNPs with BMP2, which has been shown to induce neuronal differentiation through Smad5 mediated mechanisms (Rios et al., 2004), increased rpS6 phosphorylation while decreasing CGNP proliferation (data not shown).

Figure 5.

Increased S6K activity in proliferating CGNPs promotes cell cycle exit. (A) Western blot analysis of CGNPs treated with bFGF which promotes CGNP cell cycle exit. S6 kinase activity increases while the proliferation markers cyclin D2 and Nmyc decrease. (B) S6 kinase overexpression decreases CGNP proliferation, as determined by cyclin D2 levels. (C to E) Immunohistochemical analysis of CGNPs infected with S6 K-expressing retroviruses. Untreated- (C) Shh-treated infected with control GFP viruses (D) or Shh-treated infected with S6K viruses (E) were incubated for 48 hours and then analyzed for proliferation by BrdU incorporation. (F) Proliferation of CGNPs was quantified by BrdU incorporation. Increasing S6 kinase activity inhibited CGNP proliferation by an average of 20%. n=3; (**p<0.003).

Because S6K activity correlated with CGNP differentiation signals, we next asked whether S6K over-expression is sufficient to force CGNPs to exit the cell cycle. We infected primary CGNP cultures with retroviruses expressing S6K and assayed proliferation by carrying out western blot analysis and measuring BrdU incorporation. As shown in Figure 5B, over-expression of S6K resulted in increased rpS6 phosphorylation and reduced IRS1 and cyclin D2 protein levels. These data lend further support to our previous report where we demonstrated that CGNPs stabilize IRS1 through S6K suppression to promote proliferation (Parathath et al., 2008). When we measured BrdU incorporation in Shh-treated S6K-infected CGNPs (Figure 5C-E), we observed a significant decrease in the number of BrdU positive cells as compared to control (GFP-infected) cells (p<0.003) (Figure 5F). The reduction of proliferating CGNPs in S6K-infected cultures cannot be attributed to an increase in apoptosis as determined by quantification of cleaved caspase-3-positive cells (data not shown). The inhibitory effects on proliferation as a result of increased S6K activity further support the hypothesis that in CGNPs, the downstream mTOR effectors, eIF4E and S6K, carry out distinct roles in regulation of progenitor cell expansion and cell cycle exit, respectively.

CGNPs regulate S6K activity through activation of a specific PP2A complex

Our data demonstrate that Shh treatment results in increased levels of eIF4E protein in CGNPs which can be explained by increased eIF4E mRNA expression; however, Shh caused decreased S6K activity, but did not reduce total S6K or rpS6 levels. We therefore sought to determine the mechanism by which S6K activity is suppressed in proliferating CGNPs. We have shown that treatment with okadaic acid (OA), a potent inhibitor of protein phosphatase 2A (PP2A) function, promotes S6K activity in Shh-treated CGNPs (Parathath et al., 2008), and we have also identified a positive relationship between Shh mitogenic signaling and PP2A activity in that PP2A de-phosphorylates N-myc, thereby increasing its stability (Sjostrom et al., 2005). Previous studies have suggested that mTOR may bind to and inhibit PP2A (Hartley and Cooper, 2002). However, these observations do not explain why 4E-BP phosphorylation is not suppressed by Shh in CGNPs, but that of S6K is.

To further investigate the involvement of PP2A in suppressing S6K activity we infected proliferating CGNPs with retroviruses carrying Small-T antigen, which is known to antagonize PP2A (Chen et al., 2004). As shown in Figure 6A, over-expression of Small-T antigen restored S6K activity and reduced levels of proliferation mediators IRS1 and cyclin D2, suggesting that PP2A may play a role in downstream mTOR signaling. We next measured the number of cells that incorporated BrdU in control or Small-T infected CGNPs. As shown in Figure 6B, Small-T infection resulted in a greater than 50% decrease in the number of BrdU positive cells, further supporting our hypothesis that PP2A activity plays a role in CGNP proliferation.

Figure 6.

(A) Western blot analysis of lysates from CGNPs infected with retroviruses expressing Small-T antigen. Small-T rescues rpS6 phosphorylation while decreasing proliferation, measured by cyclin D2 levels. (B) Effects on proliferation after control or Small-T infection measured by BrdU incorporation, n=3; (***p<0.0008). (C) Evaluation of PP2A regulatory subunit protein levels in vehicle and Shh-treated CGNPs by western blot analysis reveals that Shh specifically up-regulates the B56γ3 subunit while B56ε levels remain constant. (D) Western blot analysis demonstrating that treatment with cyclopamine (1μg/mL) for increasing periods of time reduces levels of the PP2A B56γ3 subunit as well as rescues S6 kinase activity. (E) eIF4E knockdown in Shh-treated CGNPs reduces B56γ subunit levels. (F) Knockdown of the PP2A B56γ3 subunit in Shh-treated CGNPs restores rpS6 phosphorylation but does not affect 4EBP2 phosphorylation. Loss of this PP2A B subunit also results in reduction of IRS1 and cyclin D2 levels (see also Figure S3). Loss of PP2A B56γ3 subunit reduces proliferation by an average of 18% as determined by BrdU incorporation quantification (G); n=3; (**p<0.03).

As a trimeric protein complex, PP2A contains a catalytic subunit (C), a scaffold subunit (A), and a regulatory “B” subunit, which directs PP2A to its specific substrates (Virshup and Shenolikar, 2009). We speculate that Shh signaling could regulate a “B” subunit targeting S6K as a way to specifically de-phosphorylate S6K. We therefore examined expression of specific PP2A “B” subunits in Shh-treated CGNPs. Using our primary culture system we found that the protein levels of the B56γ regulatory subunit were up-regulated in Shh-treated CGNPs (Figure 6C), while the B56ε subunit levels, which have been previously demonstrated to positively modulate Shh signaling during eye development, remained constant (Rorick et al., 2007). Additionally treatment with the Smoothened inhibitor cyclopamine (Berman et al., 2002) not only rescued S6K activity but also reduced B56γ protein levels (Figure 6D, see also Figure S3), further supporting a role for PP2A-B56γ complex in promoting proliferation, perhaps by inhibiting S6K. We next wished to ask whether Shh-mediated B56γ induction might lie downstream of eIF4E. Indeed, when we knocked down eIF4E with lentiviruses containing shRNAs targeting eIF4E we observed a decrease in B56γ protein (Figure 6E). Taken together with our previous knock-down experiments where we saw rescue of S6K activity, these data support our hypothesis that PP2A-B56γ mediates S6K suppression during CGNP proliferation; when CGNPs are cued to exit the cell cycle, the PP2A-B56γ complex is disrupted, allowing for S6K activity, destabilization of IRS1, and completion of cell cycle exit.

It has been reported that the Drosophila ortholog of this PP2A subunit mediates Drosphila S6K de-phosphorylation (Bielinski and Mumby, 2007), and that knocking down the mammalian ortholog increases S6K phosphorylation in HeLa cells (Hahn et al., 2010). Thus, we next asked whether S6K activity increases when the B56γ subunit is knocked down in proliferating CGNPs. Lentiviral-mediated knockdown of the B56γ subunit in proliferating CGNPs resulted in increased S6 kinase and rpS6 phosphorylation but no differences in 4EBP2 phosphorylation, indicating that B56γ loss specifically relieves the inhibition of S6K (Figure 6F). Moreover, B56γ knock-down resulted in decreased cyclin D2 and IRS1 levels, as well as reduced proliferation in Shh-treated CGNPs as indicated by BrdU incorporation into cycling cells (p<0.03) (Figure 6F, G). To address the specificity of the B56γ subunit in mediating the suppression of S6K activity we knocked down another subunit, B56ε and evaluated rpS6 phosphorylation. Loss of this subunit did not rescue S6K activity and had little effect on CGNP proliferation (Supplementary Figure 3). Taken together these data strongly suggest that Shh-mediated induction of the PP2A-B56γ complex downstream of eIF4E drives S6K de-phosphorylation and by extension maintains CGNP proliferation competency.

DISCUSSION

Here we have investigated how the Shh signaling pathway affects regulators of protein translation typically lying downstream of mTOR. Previously, we showed that Shh suppresses S6 kinase activity in order to stabilize IRS1, through an unknown mechanism. Here, we show that Shh signaling modulates individual mTOR effectors separately in order to maintain a proliferation-competent state. We identified a novel mode of mTOR effector regulation wherein Shh induces eIF4E in CGNPs, which we show is required for their proliferation, while simultaneously suppressing S6K activity via promoting its PP2A-mediated dephosphorylation (Figure 7). Conversely, induction of S6K activity through ectopic over-expression or PP2A inhibition results in CGNP cell cycle exit.

Figure 7.

Model depicting how Shh promotes CGNP proliferation through combined effects on Gli1/2 and the mTOR pathway components eIF4E and S6K. Shh binding to Ptc leads to Smo activation, resulting in up-regulation of Gli and N-myc. eIF4E mRNA and protein are up-regulated, up-stream of the PP2A-B56γ subunit, which directs PP2A activity towards S6 kinase. Inactivation (dephosphorylation) of S6 kinase results in IRS1 stabilization, which can also drive N-myc expression, thus creating a pro-proliferative feed-forward loop in CGNPs.

Previous work performed in cell lines demonstrates that mTOR inhibits 4EBPs and activates S6K in order to promote protein synthesis (Fingar et al., 2004; von Manteuffel et al., 1996). However immortalized cell lines do not recapitulate the normal life cycle of most cell types, especially those cells that dramatically alter their morphology upon differentiation. Using primary cultures established from post-natal mouse cerebella allows us to more accurately determine how CGNPs regulate cellular processes as these cultures have been proven to mimic in vivo conditions. We propose that Shh signaling in CGNPs preferentially activates eIF4E to promote cell cycle progression; when cued to exit the cell cycle, CGNPs utilize S6K functions. However, we note that since ectopic S6K expression or B56γ shRNA-mediated inhibition of S6K dephosphorylation did not completely eliminate CGNP proliferation, other downstream effectors of Smoothened, such as Gli, are likely to play essential roles in achieving the complete mitogenic response of CGNPs to Shh (Figure 7).

While it is thought that rpS6 is an important regulator of proliferation-associated mRNAs containing a highly structured 5’-untranslated region (UTR), our findings in neural precursors suggest that this downstream effect of S6K is not required for proliferation. Indeed, it has been shown that mice expressing knocked-in mutant non-phosphorylatable rpS6 show increased proliferation, reduced size, and enhanced protein synthesis in certain cell types (Ruvinsky et al., 2005), further supporting alternate roles for S6K activity. However, a normal biological setting wherein rpS6 activity is reduced in developing cells has not been shown. Here, we suggest that CGNPs exemplify a developmental paradigm for shifting the balance of mTOR effectors between eIF4E and rpS6, since they proliferate with a very short cell cycle (Mares et al., 1970), and these cells are very small, consisting mostly of nucleus. During differentiation, they extend neural processes, resulting in an increased cell size, where a role for S6K may become more important (Altman J. and Bayer, 1997; Bateman and McNeill, 2004; Shima et al., 1998). Furthermore other kinases known to target rpS6, such as p90RSK (Roux et al., 2007), may play a role in activating rpS6 phosphorylation to promote cap dependent translation while S6K mediated rpS6 phosphorylation leads to CGNP cell cycle exit.

Our data indicate an important role for eIF4E in Shh-mediated CGNP proliferation. A critical and rate-limiting component of the initiation complex, eIF4E binds to the cap-structure on mRNAs and helps recruit other initiation factors such as the scaffolding protein, eIF4G and an RNA helicase, eIF4A, to the mRNA. The cap structure consists of a 5’ methyl-7-guanosine linked to the initial nucleotide of the mRNA molecule and this cap functions to target the message for translation as well as protect it from enzymatic degradation. Assembly of the eIF4F complex on cellular mRNAs initiates the complex process of protein translation, which includes mRNA unwinding, initiation codon scanning as well as ribosome assembly. Response to growth factor signals regulates activity of eIF4E but generally the overall levels of eIF4E remain limited in cells and are kept inactive by translation inhibitor proteins; therefore capped mRNAs must compete with the limited amount of eIF4E available in order to be translated.

We find that during the rapid expansion phase of cerebellum development, Shh signaling induces expression and promotes increased protein levels of eIF4E as well as other components of the translation initiation machinery, including eIF4G. eIF4E not only co-localizes with proliferation markers in vitro but is also expressed in proliferating precursor cells of the EGL in vivo. The eIF4E promoter is known to contain myc-binding sites and increased levels of cmyc have been shown to induce eIF4E expression (Rosenwald et al., 1993). Shh signaling induces Nmyc, which promotes expression of G1 cyclins and may provide a mechanism for the induction of eIF4E (Ciemerych et al., 2002; Kenney et al., 2003; Oliver et al., 2003).

Based on the lines of evidence presented here, we propose that Shh signaling in CGNPs preferentially activates eIF4E while simultaneously inhibiting S6K activity through the activity of PP2A-γ56γ, thereby promoting cell cycle progression at least in part through stabilization of IRS1, which we have shown drives expression of N-myc and cyclin D2 (Parathath et al., 2008). While it has been shown that ectopic eIF4E expression in cell lines can impair S6K activity (Khaleghpour et al., 1999), the mechanism underlying this phenomenon was not determined, and a relevant biological context has not previously been identified. Thus, our results demonstrate a novel mechanism wherein eIF4E and S6K are differentially regulated in proliferating cerebellar neural precursors by Shh, to ultimately promote proliferation in primary neural precursor cells. Further, they suggest that eIF4E inactivation may be an effective way to reduce proliferation in CGNP-derived medulloblastomas.

MATERIALS AND METHODS

Mice

Harvest of cerebellar neural precursors and cerebella preparation for histological analysis from murine neonates were carried out in compliance with the Memorial-Sloan Kettering Institutional animal care and use committee guidelines.

Cerebellar Granule Precursor Cell Cultures

Cerebella were isolated from PN day 5 Swiss-Webster (SW) mice and primary cultures were prepared as described (Parathath et al., 2008). Recombinant Shh (R&D Systems, concentration 3μg/mL) was used to maintain proliferation as indicated. Where applicable, cyclopamine (1 μg/ml, kind gift of Dale Gardner, USDA), rapamycin (10nM, Sigma) or bFGF (20 ng/mL, Peprotech) were added for indicated times after 24 hours of Shh treatment.

Protein Preparation and Western Blotting

Adherent and floating cells were collected, washed once with PBS, then resuspended in lysis buffer and processed as previously described (Kenney and Rowitch, 2000). For each sample 30 μg were separated on 8% or 10% SDS-polyacrylamide gels and then transferred to activated PVDF membranes (Millipore). Western blotting was carried out according to standard protocols. Primary antibodies included eIF4E, phospho-eIF4E, eIF4G, phospho-eIF4G, 4EBP2, phospho-4EBP2, rpS6, phospho-rpS6 (235/236 or 240/244), mTOR, S6 kinase, phospho-S6 kinase (Cell Signaling), Raptor, Rictor (Bethyl Laboratories), cyclin D2, Nmyc (Santa Cruz Biotechnology), GFP (Invitrogen), HA (Chemicon), PP2A B56γ (Abcam), PP2A B56ε (Novus Biologicals), or β-tubulin (Sigma). Horseradish-peroxidase conjugated secondary antibodies were anti-mouse (Jackson labs) or anti-rabbit (Pierce). Blots were developed using Amersham ECL kits. Chemiluminescence was detected by exposing membranes to Kodak biomax film for various intervals to obtain a non-saturated image.

Immunoprecipitation

One mg of protein extract was used from either vehicle-treated or Shh-treated CGNP cultures. Ten micrograms of antibody were incubated with Protein A-Sepharose beads (Invitrogen) for two hours. Protein extracts were precleared with Protein A-Sepharose beads for two hours, and then incubated with the antibody plus Protein A-Sepharose solution overnight at 4°C. The precipitate was washed four times and proteins were eluted with 0.2 M glycine. Antibodies used for immunoprecipitation were eIF4E (Santa Cruz Biotechnologies), eIF4G (Cell Signaling), and mouse and rabbit IgG (Upstate Biotechnologies).

Virus Production and Infection

Lentiviruses were produced as previously described (Parathath et al., 2008). Briefly 293 EBNA (Invitrogen) packaging cells were co-transfected with lentiviral constructs (Mission shRNA, Sigma) expressing short hairpin RNAs targeting eIF4E, PP2R5c, PP2R5e or GFP, delta 8.9, and vesicular stomatitis virus G glycoprotein (vsvg) plasmids, using Fugene 6 transfection reagent (Roche). The media was changed 12 hours after transfection and supernatants (10 mL) were harvested every 24 hours for 72 hours and kept at 4°C, then pooled, filtered through 0.45 μm syringe filters, concentrated by centrifugation, and stored at 4°C until use. CGNPs were infected with 50 μL concentrated viral supernatant and maintained in N2 media with Shh. For retroviral production HA-S6K, provided by John Blenis (Harvard), and Small-T, provided by William Hahn (Harvard) were cloned into pWZL IRES-GFP vector and co-transfected with gagpol and vsvg into 293E packaging cells. Collections and infections were carried out similar to lentivirus procedures.

Immunohistochemistry and Immunofluorescence

Paraffin-embedded sections (5 μM) were de-waxed, rehydrated and fixed in ice-cold acetone for 20 minutes, then boiled in 0.01 M citric acid for 15 minutes for antigen retrieval. Sections were blocked with 10% normal goat serum (Sigma) with 0.1% Triton X-100/PBS. Primary antibodies included eIF4E, phospho-S6 (235/236), GFAP, NeuN (Cell Signaling), MEF2D, p27 (BD-PharMingen), BrdU (Becton Dickinson), and PCNA (Calbiochem). After washing in PBS, slides were incubated with either goat anti-rabbit or goat anti-mouse fluorescently tagged secondary antibodies (Invitrogen). Sections were mounted using Vectashield mounting media with DAPI (Vector Laboratories). For immunohistochemical DAB detection an automated staining processor was used (Discovery, Ventana Medical Systems, Inc.).

Quantitative PCR

RNA from CGNPs was isolated using the TRIZOL (Invitrogen) reagent and resuspended in 35γL DEPC-treated water. cDNA was generated with SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen). TaqMan Gene Expression Arrays (Applied Biosystems) using TaqMan custom designed MGB probes for eIF4E (Mm01317468_m1) and β-actin (Mm01191484_m1) were performed in triplicate on an ABI 7000 Sequence Detection System. Data was analyzed with ABI GeneAmp SDS software (Applied Biosystems). The average threshold cycle (C_T) was determined to quantify transcript levels, normalized against β-actin and the results reported as fold changes.

Image Capturing

Immunostaining performed on cultured CGNPs or tissue sections was visualized using a Leica DM5000B microscope and images were captured with Leica FW400 software. For quantification of BrdU incorporation into primary cells, TIFF images of four random fields were taken for each experimental group using the 10X objective. The percentage of BrdU-positive cells over the total number of cells, as determined by DAPI staining, was calculated using Image Pro Plus software (MediaCybernetics).

Statistics

Statistical analysis was performed using one-way ANOVA followed by a two-tailed t-test for comparisons between groups. All results are given as mean±s.e.m. All in vitro experiments were performed at least three times with separate litters to confirm reproducibility and consistency.