CREB signaling activity correlates with differentiation and survival in medulloblastoma

While there has been significant progress in the molecular characterization of the childhood brain cancer medulloblastoma, the tumor proteome remains less explored. However, it is important to obtain a complete understanding of medulloblastoma protein biology, since interactions between proteins represent potential new drug targets. Using previously generated phosphoprotein signaling-profiles of a large cohort of primary medulloblastoma, we discovered that phosphorylation of transcription factor CREB strongly correlates with medulloblastoma survival and associates with a differentiation phenotype. We further found that during normal cerebellar development, phosphorylated CREB was selectively expressed in differentiating cerebellar granule neuron progenitor (CGNP) cells. In line, we observed increased differentiation in CGNPs treated with Forskolin, Bmp6 and Bmp12 (Gdf7), which induce CREB phosphorylation. Lastly, we demonstrated that inducing CREB activation via PKA-mediated CREB signaling, but not Bmp/MEK/ERK mediated signalling, enhances medulloblastoma cell sensitivity to chemotherapy.


Scientific Reports
| (2021) 11:16077 | https://doi.org/10.1038/s41598-021-95381-0 www.nature.com/scientificreports/ and EP300 genes, respectively) 8,9 . Subsequently, the CREB-CREBBP/p300 transcriptional complex activates CRE-elements of target genes 10 . In addition, several other kinases can target CREB non-canonically such as calmodulin-dependent kinases II/IV, pp90 rsk , MSK-1, MEK/ERK and AKT, underscoring the complexity of CREB-CREBBP/p300 activity 5,11,12 . Interestingly, CREB plays a role in the differentiation of cerebellar granule neuron progenitors (CGNPs), a major cerebellar progenitor cell type that is believed to be cell-of-origin for SHH and Group 3 medulloblastoma 6,[13][14][15][16][17] . Immature CGNPs do not express phospho-CREB. However, when they initiate differentiation into mature granule neurons, active CREB levels increase 6 . Therefore, we speculate that tumors exhibiting relatively high pCREB Ser 133 levels have a more differentiated phenotype, which is frequently associated with increased sensitivity to therapy and thus improved outcome 18 . In line, we found enrichment for differentiationrelated processes in tumors expressing high pCREB Ser 133 . This prompted us to consider CREB modulation as a therapeutic strategy and explore different approaches to induce CREB phosphorylation. We found that PKAmediated CREB signaling, but not Bmp/MEK/ERK mediated signalling, enhances medulloblastoma cell sensitivity to Etoposide, suggesting that canonical CREB pathway activation may increase the impact of chemotherapy. Murine neural progenitor cells (NPCs) were isolated from dissected neonatal C57/BL6 subventricular zone as described previously 19 . NPCs were maintained as monolayers in culture media containing DMEM/F12 supplemented with 1% N2 (Life Technologies), human EGF and basic-FGF (20 ng/mL) (PeproTech), and with addition of Laminin (1 μg/mL) (Sigma). NPCs were seeded at a density of 50.000 cells/well into 12 wells plates (Greiner).
All experiments involving animal material were approved by the Institutional Animal Care and Use Committee of the University Medical Center Groningen, the Netherlands, in adherence with the ARRIVE guidelines 21 .
NPC differentiation was induced using DMEM/F12 supplemented with 2% serum in the presence or absence of CREB blocker Naphtol AS-E phosphate/KG-501 (3 μM) (Sigma) or DMSO as control for 48 h, followed by immunofluorescence staining for differentiation markers.
CGNP differentiation was induced by treatment with FSK (5 μM) (Sigma), Bmp6 or Bmp12 (PeproTech) in CGNP culture media. CGNPs were fixed at 1 day in vitro (DIV1) for mRNA isolation and subsequent quantitative RT-PCR, or at DIV4 for immunofluorescence staining (see Supplementary file).
Western blotting. Cells were lysed in RIPA buffer supplemented with cOmplete and PhosSTOP (Roche).
Survival curves. Average levels of pCREB Ser 133 peptide phosphorylation, and average levels of CREBBP and EP300 mRNA expression, were calculated for the complete cohort to determine a threshold: all samples with phosphorylation or mRNA expression levels lower or higher than the threshold were assigned as Below average (dashed blue lines) or Above average (solid red lines), respectively. For molecular subgroup analysis, the average levels of pCREB Ser 133 peptide phosphorylation were calculated for each subgroup. Kaplan-Meier curves were generated based on below (or above) pCREB Ser 133 peptide phosphorylation level and mRNA expression level for overall survival, and p-values were calculated using the log-rank (Mantel-Cox) test. All Kaplan-Meier curves and log-rank tests were generated using Prism 8 (GraphPad). For medulloblastoma primary samples, peptide phosphorylation arrays and gene expression profiling, see Supplementary info. Statistical analysis. Prism 8 (GraphPad) was used to prepare the charts and perform statistical analyses.
Ethics approval. All applicable national and institutional guidelines for the care and use of animals were followed. All animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Groningen (IACUC-RUG) (IvD 17813-01-001).

CREB activity correlates with overall survival in pediatric medulloblastoma patients.
To identify phosphoproteins correlating with medulloblastoma outcome, we analyzed the relationship between individual phosphorylated peptides from our previously generated medulloblastoma phosphoprotein-signaling profiles, with overall survival 4 . We found that patients with relatively low levels of the peptide representing the phosphorylated (active) form of the transcription factor CREB (pCREB Ser 133 ) had an extremely poor prognosis, indicating that high CREB activity correlates with overall survival (Fig. 1a). Interestingly, such relationship was not found when comparing CREB1 mRNA levels with survival ( Fig. 1b), suggesting that the role of CREB in medulloblastoma is confined to the protein compartment.
We next questioned if the correlation between pCREB Ser 133 levels and survival was subgroup specific. Hereto, we split the cohort by molecular subgroup, which notably leads to relatively small samples sizes (SHH, n = 13; Group 3, n = 15; Group 4, n = 17) 23 . We found that levels of pCREB Ser 133 were particularly predictive for outcome in SHH and Group 3 medulloblastoma, whereas in Group 4 medulloblastoma only a trend was observed ( Fig. 1c-e). Strikingly, relative pCREB Ser 133 peptide phosphorylation and CREB1 mRNA levels were highest in Group 4 patients ( Supplementary Fig. S1), suggesting that pCREB Ser 133 levels are predictors of outcome in specific contexts. Together, these results suggest an important role for CREB activity in the prognosis of medulloblastoma.
The CREB transcriptional complex is linked to medulloblastoma survival. Phosphorylation of CREB Serine 133 induces binding to the histone acetyltransferase CREBBP or its homolog p300, to form CREB/ CREBBP or CREB/p300 transcriptional complexes 8,9 . To explore a role for CREB-complexes in medulloblastoma, we first investigated the relationship between CREBBP and EP300 mRNA expression and overall survival. We found that patients expressing higher CREBBP and EP300 mRNA expression had improved outcome, and this was significant for EP300 (p = 0.0111) (Fig. 2a,b; Supplementary Fig. S2). We then grouped the medulloblastoma samples into below (or above) average pCREB Ser 133 (pCREB Ser 133 hi/lo ), in combination with below (or above) average CREBBP (CREBBP hi/lo ) or EP300 (EP300 hi/lo ) mRNA levels, and assessed their relationship with overall survival. Samples that did not meet these criteria were arbitrarily assigned to the Intermediate group. We found no difference in overall survival between patients from the Intermediate group and the pCREB Ser 133/hi CREBBP/EP300 hi groups (Fig. 2c,d). Remarkably, we did observe highly significant correlations between poor outcome and low pCREB Ser 133 levels in combination with either low CREBBP (p = 0.0013), or EP300 (p < 0.0001) mRNA expression (Fig. 2c,d).
We then functionally confirmed that CREB complex members impact on medulloblastoma cell survival. Hereto, we performed knockdown of either CREB1, CREBBP, or EP300 in MED8A cells, an in vitro model for Group 3 medulloblastoma (Fig. 2e) and subsequently determined their sensitivity to the chemotherapeutic agent Etoposide. In agreement with our findings in patients, we observed that loss of CREB-complex expression     www.nature.com/scientificreports/ increased medulloblastoma cell viability (Fig. 2f). These data suggest that CREB/p300, and to a lesser extent CREB/CREBBP, are beneficial for medulloblastoma outcome.

CREB-complex activity is associated with neural differentiation in tumors and developing
CGNPs. It is known that CREB-complex activity is associated with neural differentiation, including in the cerebellum 6,13 . We therefore questioned if the medulloblastoma samples in our cohort with high CREB-complex activity also exhibited a differentiated phenotype, which could explain the improved outcome 18 . We again categorized the medulloblastoma samples into three groups based upon pCREB Ser 133 hi/lo levels, in combination with CREBBP hi/lo and EP300 hi/lo . We then took advantage of our data set of medulloblastoma transcriptomes that we had generated previously 4 , and performed supervised hierarchical clustering according to the three groups. This revealed two robust gene expression clusters that were reciprocally enriched in the pCREB Ser 133 hi /CREBBP hi / EP300 hi (black) or pCREB Ser 133 lo /CREBBP lo /EP300 lo (white) groups (Fig. 3a). Samples from the Intermediate group (grey) exhibited either gene expression profile. Interestingly, subsequent functional annotation of these two gene clusters revealed that biological processes related to neuronal differentiation and development were highly enriched in the pCREB Ser 133 hi /CREBBP hi /EP300 hi group (Fig. 3b). No enriched biological processes were found in genes enriched in the pCREB Ser 133lo /CREBBP lo /EP300 lo group. Of note, when we specifically analyzed the expression levels of genes involved in CGNP differentiation (i.e., ZIC and GRIN genes) across the medulloblastoma cohort, we found that ZIC2, GRIN2A and GRIN2C were relatively high in SHH tumors, even though these tumors clustered mostly to the Intermediate and pCREB Ser 133lo /CREBBP lo /EP300 lo group ( Supplementary  Fig. S3). This suggests that not all differentiation is blocked in pCREB Ser 133lo /CREBBP lo /EP300 lo tumors.
To further explore the link between the CREB complex and differentiation, we next functionally tested if the interaction between CREB and CREBBP/p300 is required for neural differentiation using neural progenitor cells (NPCs) as an in vitro model. We induced NPC differentiation in the presence or absence of a low dose of KG501, a blocker of the interaction between CREB and CREBBP/p300 24 , and found that while NPCs rapidly induce expression of differentiation marker Gfap and lose expression of stem cell marker Nestin in the presence of serum, KG-501 treatment completely inhibits the appearance of differentiated cells (Fig. 3c). This indicates that the CREB-CREBBP/p300 interaction is required for normal NPC differentiation.
We subsequently investigated phospho-CREB expression in the putative cell-of-origin for SHH and Group 3 medulloblastoma, which is the CGNP 14,16,17 . CGNPs reside in the external granular layer (EGL) of the developing cerebellum before migrating to the internal granular layer (IGL), where they mature into granule neurons 25,26 . We examined the expression pattern of CREB and phospho-CREB in these cells in postnatal day 7 (P7) and day 30 (P30) cerebellum (Fig. 3d,e). In line with earlier findings 27 , we find that total CREB is expressed in the majority of CGNPs in all cellular layers (Fig. 3d). In contrast, phospho-CREB is selectively expressed in areas enriched for differentiating CGNPs, including the P7 inner EGL (iEGL) and IGL (Fig. 3e). The outer EGL (oEGL), where immature proliferating cells reside, is negative for phospho-CREB. These data suggest a role for active CREB in CGNP differentiation.

CREB pathway activation induces CGNP differentiation. We then wondered if inducing CREB
pathway activation would result in CREB phosphorylation and subsequent differentiation. We focused on the PKA pathway that can be stimulated by the compound Forskolin (FSK), and the MEK/ERK pathway that is responsive to Bmp (Bone Morphogenetic Protein) factors (Fig. 4a). We selected Bmp6 and Bmp12 (Gdf7) as ligands for Bmp pathway stimulation, since we found that they are overlappingly expressed with phospho-CREB at sites of CGNP differentiation (Fig. 4b) 28 .
We first tested if FSK/Bmp6/Bmp12 were able to induce phospho-CREB expression in primary CGNP cultures. We found that all three compounds could induce phospho-CREB, with FSK being the most potent (Fig. 4c). We further found that FSK/Bmp6/Bmp12 treatment induced CGNP differentiation, as shown by an increase in Doublecortin (Dcx) positive neurites and connections between cells (Fig. 4d). Moreover, we observed a significant increase in mRNA expression of differentiation markers Dcx and glutamate receptor subunit Grin1 (Fig. 4e). Resuming, all three compounds induce phospho-CREB and promote CGNP differentiation, albeit to different extents, with Forskolin being the most potent except for the formation of Dcx neurites.

Exploiting CREB pathway activation to target medulloblastoma cells.
We then wanted to explore if FSK/Bmp treatment can enhance medulloblastoma chemosensitivity. Since it has been suggested that FSK and Bmp6/12 pathways are (partially) non-redundant, which could translate into synergistic effects upon co-treatment, we first studied pathway usage by FSK and Bmp6/12. Hereto, we employed a cell culture model (i.e., RPE-1 cells) resembling normal, non-transformed human cells, to study the pathways under native conditions [29][30][31] . This allowed us to study CREB activity in normal, unaltered signaling pathways. We first confirmed that treatment of serum starved RPE-1 cells with FSK, Bmp6, and Bmp12 was able to induce phospho-CREB expression (Fig. 5a). Simultaneous inhibition of the PKA pathway using PKA inhibitor H89 resulted in a reduction in phospho-CREB expression in all cases, suggesting that PKA is involved in relaying both FSK and Bmp signals towards CREB (Fig. 5b, Supplementary Fig. S4a). However, while inhibition of the MEK/ERK pathway using PD98059 resulted in a significant phospho-CREB reduction after Bmp6 treatment, we only observed a trend towards inhibition for Bmp12 and FSK (Fig. 5c, Supplementary Fig. S4b,c). Unexpectedly, we found that FSK treatment also led to a reduction in phospho-ERK signal regardless of PD98059 treatment, suggesting further crosstalk and partial redundancy between the two pathways.
Finally, to test if FSK/Bmp6/Bmp12 can enhance the response to chemotherapy, we co-treated RPE-1 cells with Etoposide and FSK, Bmp6, or Bmp12. We found that cell viability after treatment with Etoposide and FSK is lower than after Etoposide treatment alone, whereas Bmp6/Bmp12 treatment had no effect (Fig. 5d)   www.nature.com/scientificreports/ we observed no synergistic effect from co-treatment with FSK and Bmps. We then repeated these experiments in SHH (UW426) and Group 3 (MED8A) medulloblastoma cells and observed that both cell lines responded to FSK/Bmp6/Bmp12 treatment with CREB phosphorylation following serum starvation, albeit at varying rates (Fig. 5e). Similar to our observations in RPE1 cells, we also found that FSK treatment increased chemosensitivity (Fig. 5f). Altogether, our results show that while FSK/Bmp6/Bmp12 are all able to induce CREB phosphorylation, only FSK-mediated CREB activation might be useful as chemotherapy adjuvant.

Discussion
By investigating the medulloblastoma phosphoproteome, we identified relatively high levels of active (phosphorylated) CREB as a strong predictor for improved overall survival. We further found that processes related to neuronal differentiation are enriched in tumors with high CREB complex activity. This prompted us to explore the link between CREB phosphorylation and differentiation, which we hypothesized could be exploited for medulloblastoma treatment. Intriguingly, while we found that pCREB Ser 133 levels were highest across Group 4 patients, they were most significantly associated with outcome in SHH and Group 3 patients. This suggests that CREB phosphorylation per se is not sufficient to determine outcome, but that context is important. Of note, in a study investigating alternative splicing in medulloblastoma, it was found that CREB signaling-related processes were enriched in SHH and Group 3, but not Group 4 medulloblastoma 32 . Thus, it is plausible that SHH and Group 3 medulloblastoma are intrinsically more dependent on CREB signaling and as a result have increased sensitivity to alterations in CREB activity. This may be related to the fact that SHH and Group 3 medulloblastoma originate from the same cellular compartment, as further discussed below.
CREB-complex activity is associated with cerebellar development. The CREB-complex can induce transcription of a large number of CRE-containing genes 5 . The precise set of target genes is context dependent, implying that the overall outcome of CREB activation varies between cell types. This is exemplified by the fact that whereas several previous studies, including our own, have associated active CREB with pro-oncogenic functions, for medulloblastoma we now report the opposite [33][34][35] . This could originate from the neuronal context of medulloblastoma. It is known that in neural differentiation, CREB is a key driving factor and in cerebellum, CREB is associated with CGNP differentiation 6,13 . Intriguingly, these CGNPs are putative cells-of-origin for SHH and Group 3 medulloblastoma, which aligns with our finding that SHH and Group 3 survival most prominently correlates with CREB activity [14][15][16][17] . Early CGNPs undergo intense proliferation in the EGL, a secondary germinal zone 25,26 . Upon reaching maturity, they migrate inwards to form the definitive IGL of granule neurons 25 . We speculate that the CREB-complex controls this process and that its activity prevents the formation of (pre)cancerous lesions 36 . This is in agreement with observations in a Crebbp null mouse model, in which early postnatal Crebbp loss enhances medulloblastoma formation 37 . Further, patients with Rubinstein-Taybi syndrome have an increased risk of developing medulloblastoma, and CREBBP mutations have been found in sporadic medulloblastoma [38][39][40] . Following from this, we reasoned that inducing CREB phosphorylation may be beneficial in medulloblastoma therapy, and set out to identify compounds that can activate CREB in cerebellar cells. It had been reported that CREB activation is linked to extracellular factors, including Bmp family members, in areas of CGNP differentiation 25,28,41 . In addition to Bmp6, we focused on Bmp12 (Gdf7), which is a powerful player in early CGNP specification that has not yet been studied in the context of CGNP differentiation 42 . Although we discovered that CGNPs do not express Bmp6/12 themselves, we could detect Bmp6 and Bmp12 in the direct vicinity of differentiating CGNPs expressing high phospho-CREB. This suggests that other cerebellar cells supply these factors to neighboring CGNPs. Indeed, using primary cerebellar cell cultures, we could demonstrate that ectopic Bmps induce CREB phosphorylation and CGNP differentiation, and therefore are potential therapy candidates.
Modulating CREB-complex activity. When we started to explore CREB activation as potential medulloblastoma treatment, alongside Bmps that are believed to activate CREB via MEK/ERK signaling, we included Forskolin (FSK), a potent activator of the cAMP/PKA pathway also targeting CREB 43,44 . We reasoned that targeting CREB via alternative pathways might have a synergistic effect on tumor cell chemosensitivity. However, we found that PKA inhibition also had a negative effect on Bmp-induced CREB activation, highlighting a role for PKA in the Bmp-CREB axis 45,46 . Further, we observed a decrease in MEK/ERK phosphorylation after FSK treatment 47,48 . Thus, the Bmp and FSK pathways are partially redundant and in line with this, we found no synergistic effect of Bmp and FSK co-treatment on medulloblastoma cell chemosensitivity. In fact, while FSK treatment increased chemosensitivity in all cell lines tested, for Bmp we only found such effect in UW426 cells, possibly because Bmps induce less efficient CREB phosphorylation.
Altogether, we conclude that inducing CREB activation through the PKA but not Bmp signaling cascade might be beneficial as chemotherapy adjuvant in medulloblastoma. Forskolin seems to be a promising compound for this, especially since this drug has already been clinically tested as treatment for several human diseases including diabetes and liver fibrosis; and there is increasing interest in Forskolin as anti-cancer treatment 44,49 .

Data availability
All data generated for this study are included in this manuscript or can be found at Mendeley Data, http:// dx. doi. org/ 10. 17632/ s7j3g kd58p.1.